Laser ultrasonic prediction method for grain size of steel plate

Abstract

The invention relates to the technical field of metal material grain size detection, in particular to a laser ultrasonic prediction method for the grain size of a steel plate, and the method comprises the steps: obtaining a laser ultrasonic grain size detection process, carrying out the signal feature recognition and extraction, and obtaining a plurality of ultrasonic feature impact factors; aiming at each ultrasonic characteristic influence factor, collecting laser ultrasonic signal data in multiple experiments; on the basis of a pre-constructed feature quantification mechanism, performing feature evaluation on all the laser ultrasonic signal data corresponding to the ultrasonic feature influence factor to obtain a quantitative evaluation value of the ultrasonic feature influence factor; carrying out data fusion on the plurality of quantitative evaluation values, and constructing a grain size characterization vector; and inputting the grain size characterization vector into a pre-constructed grain size inversion model to obtain a grain size prediction result of the metal material to be measured. Quantitative lossless characterization of the average grain size of the metal plate can be achieved, and the detection precision and robustness are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of metal material grain size detection, and in particular to a laser ultrasonic prediction method for steel plate grain size. Background Art

[0002] Metals and alloy materials are processed into various products due to their excellent mechanical properties and are indispensable materials in our lives. In the study of alloy materials, it was found that the material is composed of countless grains. Small grains can produce better mechanical properties, so this is a common way to improve the performance of alloys. Many researchers have studied the relationship between grain size and mechanical properties, named the Hall-Petch effect. Specifically, fine-grained structure improves strength and toughness through the Hall-Petch effect, while coarse grains are prone to high-temperature creep failure and local mechanical weaknesses. Therefore, grain size needs to be detected online during the industrial production and service of metal materials.

[0003] The most commonly used methods for measuring the grain size of metal materials in industrial settings are metallography and ultrasonic methods. Metallography, which involves etching the sample's grain boundaries and imaging them with an optical microscope, is intuitive but highly destructive, requires cumbersome sample preparation, and lacks local representation. Its limited detection range prevents online application. Traditional ultrasonic methods, while non-destructive, are limited by millimeter-level spatial resolution, couplant dependency, and model inversion errors, making it difficult to accurately characterize micron-level grain structures. Summary of the Invention

[0004] The present invention provides a laser ultrasonic prediction method for steel plate grain size, which can achieve quantitative non-destructive characterization of the average grain size of metal plates and improve the accuracy and robustness of detection, and can effectively solve the problems in the background technology.

[0005] To achieve the above objectives, in a first aspect, the present invention provides a laser ultrasonic prediction method for steel plate grain size, comprising: Obtain the laser ultrasonic grain size detection process, perform signal feature recognition and extraction on it, and obtain multiple ultrasonic feature influencing factors; For each of the ultrasonic characteristic influencing factors, collecting laser ultrasonic signal data from multiple experiments; Based on a pre-established feature quantification mechanism, feature evaluation is performed on all laser ultrasonic signal data corresponding to the ultrasonic feature influencing factor to obtain a quantitative evaluation value of the ultrasonic feature influencing factor; Performing data fusion on a plurality of the quantitative evaluation values to construct a grain size characterization vector; The grain size characterization vector is input into a pre-built grain size inversion model to obtain a grain size prediction result of the metal material to be tested.

[0006] In combination with the first aspect, in one possible design, the laser ultrasonic signal data includes time domain signal data, frequency domain signal data, and spatial distribution signal data.

[0007] In combination with the first aspect, in one possible design, the key time domain parameters of the time domain signal data include waveform, amplitude, and arrival time.

[0008] In combination with the first aspect, in one possible design, the key frequency domain parameters of the frequency domain signal data include.

[0009] In combination with the first aspect, in one possible design, key parameters of the spatially distributed signal data include signal strength distribution and phase information.

[0010] In combination with the first aspect, in a possible design, factors influencing the construction of the characteristic quantization mechanism include metal type, alloy composition, ultrasonic signal type, ultrasonic signal characteristic stability, laser parameters, and environmental factors.

[0011] In conjunction with the first aspect, in one possible design, based on a pre-established feature quantization mechanism, feature evaluation is performed on all signal data corresponding to the ultrasound feature influencing factor, and obtaining a quantitative evaluation value of the ultrasound feature influencing factor includes: For each ultrasound feature influencing factor, a corresponding feature quantification mechanism is pre-built; Preprocessing the collected laser ultrasonic signal data; According to the pre-built feature quantification mechanism, all signal data corresponding to each ultrasonic feature influencing factor are calculated to obtain the quantitative evaluation value of the feature influencing factor; Verify the calculated quantitative evaluation values.

[0012] In combination with the first aspect, in a possible design, the grain size characterization vector is input into a pre-built grain size inversion model to obtain a grain size prediction result of the metal material to be tested: Collect metal material samples with known grain sizes and obtain the grain size characterization vector corresponding to each sample; Selecting a model algorithm based on the relationship between the grain size characterization vector and the actual grain size; the model algorithm includes linear regression, support vector machine and artificial neural network; Use the training set data to train the selected model algorithm and adjust the model parameters to make the model accurately fit the training data; During the training process, the cross-validation method is used to evaluate the performance of the model, and the model is optimized based on the evaluation results; Performing preprocessing operations on the grain size characterization vector, including normalization and standardization; The preprocessed grain size characterization vector is input into a pre-built grain size inversion model to output the grain size prediction result of the metal material to be tested.

[0013] On the other hand, the present application provides an electronic device comprising a bus, a transceiver, a memory, a processor, and a computer program stored in the memory and runnable on the processor, wherein the transceiver, the memory, and the processor are connected via the bus, and the computer program implements the steps of any one of the above methods when executed by the processor.

[0014] In a third aspect, the present application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in any one of the above methods. The technical solution of the present invention can achieve the following technical effects: Compared with the metallographic method, the laser ultrasonic prediction method does not rely on corrosion treatment of the sample, so it will not cause damage to the sample and is suitable for the online detection needs of metal materials in industrial sites; Traditional ultrasonic methods are limited by millimeter-level spatial resolution and are difficult to accurately characterize micron-level grain structures. Laser ultrasonic technology, on the other hand, may have higher spatial resolution and can more accurately detect the size of micron-level grains. Traditional ultrasonic methods typically require the use of coupling agents to transmit ultrasonic waves, which not only increases the complexity of detection but also may introduce additional errors. Laser ultrasonic technology can reduce or eliminate the reliance on coupling agents, thereby simplifying the detection process and improving detection accuracy. This method extracts multiple ultrasonic feature influencing factors through signal feature recognition, evaluates and quantifies the characteristics of each factor, and finally constructs a grain size characterization vector through data fusion. This helps to more comprehensively reflect the actual grain size and improve the accuracy of prediction. By building a grain size inversion model and inputting the grain size characterization vector into the model, the grain size prediction result of the metal material to be tested can be obtained; this can process the test data more scientifically and systematically, reduce the interference of human factors, and improve the reliability and consistency of the test; Due to the advantages of laser ultrasonic prediction method such as non-destructiveness, high spatial resolution and reduced dependence on coupling agents, it is more suitable for online detection of metal materials in industrial sites; it helps to promptly discover problems in the production process, adjust process parameters, and improve product quality and production efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0016] Figure 1 Flowchart of the laser ultrasonic prediction method for steel plate grain size. DETAILED DESCRIPTION

[0017] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0018] The present application is described below in conjunction with the accompanying drawings.

[0019] like Figure 1 As shown, a laser ultrasonic prediction method for steel plate grain size of the present invention specifically includes the following steps: S1. Obtain a laser ultrasonic grain size detection process, perform signal feature recognition and extraction on it, and obtain multiple ultrasonic feature influencing factors; S2. For each of the ultrasonic feature influencing factors, collecting laser ultrasonic signal data from multiple experiments; S3. Based on a pre-established feature quantization mechanism, perform feature evaluation on all laser ultrasonic signal data corresponding to the ultrasonic feature influencing factor to obtain a quantitative evaluation value of the ultrasonic feature influencing factor; each ultrasonic feature influencing factor corresponds to a feature quantization mechanism; S4, performing data fusion on the plurality of quantitative evaluation values to construct a grain size characterization vector; S5. Input the grain size characterization vector into a pre-built grain size inversion model to obtain a grain size prediction result of the metal material to be measured.

[0020] In this embodiment, compared with the metallographic method, the laser ultrasonic prediction method does not rely on corrosion treatment of the sample, so it does not cause damage to the sample and is suitable for the online detection needs of metal materials in industrial sites. Traditional ultrasonic methods are limited by millimeter-level spatial resolution and are difficult to accurately characterize micron-level grain structures. Laser ultrasonic technology, on the other hand, may have higher spatial resolution and can more accurately detect the size of micron-level grains. Traditional ultrasonic methods typically require the use of coupling agents to transmit ultrasonic waves, which not only increases the complexity of detection but also may introduce additional errors. Laser ultrasonic technology can reduce or eliminate the reliance on coupling agents, thereby simplifying the detection process and improving detection accuracy. This method extracts multiple ultrasonic feature influencing factors through signal feature recognition, evaluates and quantifies the characteristics of each factor, and finally constructs a grain size characterization vector through data fusion. This helps to more comprehensively reflect the actual grain size and improve the accuracy of prediction. By building a grain size inversion model and inputting the grain size characterization vector into the model, the grain size prediction result of the metal material to be tested can be obtained; this can process the test data more scientifically and systematically, reduce the interference of human factors, and improve the reliability and consistency of the test; Due to the advantages of laser ultrasonic prediction method such as non-destructiveness, high spatial resolution and reduced dependence on coupling agents, it is more suitable for online detection of metal materials in industrial sites; it helps to promptly discover problems in the production process, adjust process parameters, and improve product quality and production efficiency.

[0021] In some embodiments of the present invention, with respect to step S1, The laser ultrasonic grain size detection process includes: Use high-energy laser pulses to irradiate the surface of metal materials and generate ultrasonic waves in the metal plate through thermal elastic or ablation effects; Ultrasonic signal reception: A continuous laser dual-beam heterodyne interferometer is used to receive ultrasonic signals transmitted from inside the material on opposite sides of the same axis. The ultrasonic signal contains information about the internal structure of the material. Pre-process the received ultrasonic signal, such as filtering, amplification, digitization, etc., to facilitate subsequent feature recognition and extraction; Using time domain analysis, frequency domain analysis, and time-frequency analysis, characteristic information related to grain size is extracted from the preprocessed signal. This characteristic information includes the signal's amplitude, duration, frequency components, spectral characteristics, and combined time-frequency characteristics, each of which reflects different aspects of grain size. Through experiments and data analysis, it is determined which characteristic information has a significant correlation with grain size, and then these characteristic information are used as ultrasonic characteristic influencing factors.

[0022] In this embodiment, high-energy laser pulses are used to irradiate the surface of the metal material, and ultrasonic waves are generated inside the material by utilizing the thermal elastic or ablation effect, thereby avoiding the damage to the sample by the traditional metallographic method, thereby realizing non-destructive grain size detection, which is suitable for on-line monitoring of materials in industrial production; a continuous laser double-beam heterodyne interferometer is used to receive ultrasonic signals on different sides of the coaxial axis. This configuration enhances the accuracy and sensitivity of signal reception, helps to capture tiny structural changes inside the material, and provides a high-quality data basis for subsequent feature extraction; the received ultrasonic signals are subjected to preprocessing steps such as filtering, amplification, and digitization, which effectively removes noise interference, enhances signal characteristics, creates favorable conditions for subsequent feature recognition and extraction, and improves the accuracy and reliability of feature extraction; through various methods such as time domain analysis, frequency domain analysis, and time-frequency analysis, the ultrasonic signals are preprocessed from the preprocessing The multi-dimensional characteristic information related to the grain size is extracted from the subsequent signal, including the amplitude, duration, frequency component, spectral characteristics and time-frequency joint characteristics of the signal; these characteristic information reflects the change of grain size from different angles, providing rich data support for subsequent characteristic evaluation and grain size inversion; through experiments and data analysis, it is possible to accurately determine which characteristic information has a significant correlation with the grain size, so that these characteristic information can be used as ultrasonic characteristic influencing factors; ensuring the pertinence and effectiveness of subsequent characteristic evaluation and grain size inversion, and improving the accuracy of grain size prediction; this step provides an efficient and accurate technical means for the online detection of the grain size of metal materials through non-destructive laser ultrasonic testing technology, optimized signal reception and preprocessing process, comprehensive feature extraction method and accurate feature influencing factor determination.

[0023] In some embodiments of the present invention, for step S2, In the field of metal grain size detection, in order to obtain accurate and reliable grain size prediction results, it is necessary to establish a feature quantification mechanism and grain size inversion model based on a large amount of experimental data. Sampling laser ultrasonic signal data from multiple experiments for each ultrasonic feature influencing factor can ensure the adequacy and representativeness of the data, thereby improving the accuracy of subsequent feature evaluation, data fusion, and grain size inversion. By collecting data through multiple experiments, we can obtain enough sample points to fully reflect the changing patterns of each ultrasonic feature influencing factor under different grain sizes. Multiple experiments can cover different experimental conditions, thus ensuring that the collected data is broadly representative and can reflect various situations in actual industrial production. By taking the average value or performing statistical analysis on multiple experiments, we can effectively reduce the impact of random errors on experimental results and improve data reliability. Clarify the parameter settings of the laser ultrasonic testing system, including laser energy, pulse width, repetition frequency, etc.; Determine the type, thickness, surface condition, etc. of the metal material to be tested; Set experimental conditions, such as ambient temperature and humidity, to control experimental variables; For each ultrasonic characteristic influencing factor, the variation of the factor under different grain sizes is investigated experimentally. The experimental plan includes steps such as sample preparation of different grain sizes, laser ultrasonic signal acquisition, and data recording; According to the experimental plan, each sample is tested multiple times using a laser ultrasonic testing system to collect laser ultrasonic signal data; ensuring that the conditions of each test are consistent to reduce experimental errors; Record the data of each test, including signal waveform, amplitude, duration and other information; Organize the collected laser ultrasonic signal data to remove noise and outliers; Store the organized data in a database or file; The laser ultrasonic signal data includes time domain signal data, frequency domain signal data and spatial distribution signal data; The key time domain parameters of time domain signal data include Waveform: The waveform of the laser ultrasonic signal in the time domain reflects the time course and amplitude change of the ultrasonic wave propagating in the material. The waveform contains multiple components such as direct wave, reflected wave, and scattered wave. These components are closely related to the grain structure and defects of the material. Amplitude: The amplitude of the signal indicates the intensity of the ultrasound wave and is affected by many factors, such as laser energy, material attenuation characteristics, and grain size. Changes in amplitude can reflect differences in the internal structure of the material. Arrival time: The arrival time of different ultrasonic components can be used to calculate the propagation speed of ultrasonic waves in the material, which is then related to the material's physical properties such as elastic modulus and density, which are indirectly related to grain size. The key frequency domain parameters of frequency domain signal data include: Spectrum: By performing Fourier transform on the time domain signal, the spectrum of the signal can be obtained. The spectrum contains the amplitude information of different frequency components, which are related to the grain size, defect type, etc. of the material. Center frequency: The center frequency of the spectrum can reflect the main energy distribution of the ultrasonic wave, which is related to the acoustic properties and grain structure of the material; Bandwidth: The bandwidth of the spectrum represents the frequency range contained in the signal. The change in bandwidth is related to the scattering and attenuation characteristics of the material, which are in turn affected by the grain size. Key parameters of spatially distributed signal data include: Signal intensity distribution: In laser ultrasonic testing, the signal intensity distribution at different locations on the material surface can be obtained by scanning. This distribution can reflect the spatial variation of the grain size within the material, because differences in grain size lead to different scattering and attenuation characteristics of ultrasonic waves, thus affecting the signal intensity. Phase information: In some cases, the phase information of the laser ultrasonic signal may also contain information about the grain size; the phase change of the ultrasonic wave can be obtained through interferometry technology, and this change may be related to the elastic wave propagation characteristics and grain structure of the material.

[0024] In this embodiment, by collecting data through multiple experiments, enough sample points can be obtained to fully reflect the changing law of each ultrasonic feature influencing factor under different grain sizes; at the same time, multiple experiments can cover different experimental conditions to ensure that the collected data are broadly representative and can reflect various situations in actual industrial production, laying a solid foundation for the subsequent establishment of accurate and reliable feature quantification mechanisms and grain size inversion models; multiple experiments and averaging or statistical analysis can effectively reduce the impact of random errors on experimental results and greatly improve the reliability of data; in the field of grain size detection, reliable data is the key to obtaining accurate prediction results. This step collects data through multiple experiments to ensure data quality from the source; the key time domain parameters such as waveform, amplitude, arrival time, etc. obtained can intuitively reflect the time history, intensity and propagation speed of ultrasonic waves in the material. This information is closely related to the grain structure, defects, etc. of the material, providing an important basis for subsequent feature evaluation; the key frequency domain parameters such as spectrum, center frequency, bandwidth, etc. obtained by Fourier transform can reveal the amplitude information of different frequency components in the signal, as well as the ultrasonic The main energy distribution and frequency range of the sound wave are closely related to the grain size, defect type, scattering characteristics, attenuation characteristics, etc. of the material, which helps to deeply understand the relationship between material properties and grain size; the key parameters such as the collected signal intensity distribution and phase information can reflect the spatial variation of the grain size inside the material and the propagation characteristics of elastic waves, providing strong support for a comprehensive understanding of the material grain structure from the spatial dimension; the rich and reliable laser ultrasonic signal data collected provides sufficient and high-quality samples for subsequent feature evaluation, data fusion and grain size inversion steps; in the feature evaluation stage, the influencing factors of each ultrasonic feature can be accurately quantified based on these data; in the data fusion stage, multiple quantitative evaluation values can be more effectively fused into a grain size characterization vector; in the grain size inversion stage, it helps to build a more accurate inversion model, thereby improving the accuracy and reliability of grain size prediction; this step collects laser ultrasonic signal data through multiple experiments, which plays an important role in improving data quality, comprehensively obtaining key information and assisting subsequent analysis and processing, providing a strong guarantee for the accurate prediction of the grain size of metal materials.

[0025] In some embodiments of the present invention, for step S3, Factors influencing the construction of the feature quantization mechanism include: Metal type. Different types of metals have different physical properties such as crystal structure, elastic modulus, density, etc. These properties will affect the propagation characteristics of ultrasound in the material, thereby affecting the relationship between ultrasonic characteristics and grain size; Alloy composition: The type and content of elements in the alloy affect the material's grain growth and microstructure, which in turn affects the ultrasonic signal characteristics. Adding different alloying elements to steel will change the steel's phase transition temperature and grain growth rate, leading to changes in grain size and ultrasonic characteristics. Therefore, the influence of alloy composition needs to be considered when constructing a feature quantification mechanism. Ultrasonic signal type. Laser ultrasonic signals include time-domain signals, frequency-domain signals, and spatially distributed signals. Different types of signals contain different information and have varying sensitivities to grain size. The amplitude and arrival time of the time-domain signal can reflect the propagation speed and attenuation characteristics of the ultrasonic wave, while the spectrum and center frequency of the frequency-domain signal can reflect the scattering characteristics of the material. Therefore, it is necessary to construct corresponding feature quantification mechanisms based on different signal types. Stability of ultrasonic signal features. Some ultrasonic features may be significantly affected by factors such as experimental conditions and noise, and may have poor stability. When constructing a feature quantization mechanism, it is necessary to select features with good stability or appropriately process unstable features to improve the reliability of the feature quantization mechanism. Laser parameters, such as laser energy, pulse width, and repetition frequency, affect the interaction between the laser and the material, thereby affecting the generation and propagation of ultrasonic signals. When constructing a feature quantization mechanism, the influence of laser parameters needs to be considered and the laser parameters must be kept stable during the experiment. Environmental factors, such as ambient temperature and humidity, can affect the physical properties of materials and the propagation characteristics of ultrasonic signals. Increased temperature can reduce the elastic modulus of materials and accelerate ultrasonic propagation. Therefore, environmental factors need to be controlled during the experiment, or compensation for environmental factors needs to be considered when constructing a feature quantification mechanism. For each ultrasound feature influencing factor, a corresponding feature quantification mechanism is pre-built; Before feature evaluation, the collected laser ultrasonic signal data is preprocessed, including noise removal, filtering, normalization and other operations to improve the quality and reliability of the data and ensure the accuracy of feature evaluation; According to the pre-built feature quantification mechanism, all signal data corresponding to each ultrasonic feature influencing factor are calculated to obtain the quantitative evaluation value of the feature influencing factor; The calculated quantitative evaluation value is verified by conducting comparative experiments with standard samples of known grain size, or by using other reliable detection methods. If it is found that there is a large deviation between the quantitative evaluation value and the actual situation, the characteristic quantification mechanism needs to be adjusted and optimized until the evaluation result meets the requirements.

[0026] In this embodiment, various factors that affect grain size detection are fully considered, such as metal type, alloy composition, ultrasonic signal type, feature stability, laser parameters, and environmental factors. The constructed feature quantization mechanism can more accurately reflect the complex relationship between ultrasonic features and grain size; it helps to reduce errors caused by ignoring key factors, thereby improving the accuracy of subsequent grain size prediction; the collected laser ultrasonic signal data is subjected to preprocessing operations such as noise removal, filtering, and normalization, which effectively improves the quality and reliability of the data; high-quality data is the basis for accurate feature evaluation, which can avoid the influence of interference factors such as noise on the evaluation results, and ensure that the feature quantization mechanism can be calculated based on real and effective data, and then Improve the accuracy of grain size prediction; pay attention to the stability of ultrasonic signal characteristics, select characteristics with better stability for quantitative evaluation, or properly process unstable characteristics; this enables the characteristic quantization mechanism to adapt to different experimental conditions and noise environments, improves the reliability and stability of the mechanism, and ensures that more accurate quantitative evaluation values can be obtained under different circumstances; verify the calculated quantitative evaluation values, and promptly detect deviations between the quantitative evaluation values and the actual situation by comparing them with standard samples of known grain sizes or by using other reliable detection methods; if there are deviations, the characteristic quantization mechanism can be adjusted and optimized so that the characteristic quantization mechanism can continuously adapt to the actual situation and improve its adaptability and reliability.

[0027] In some embodiments of the present invention, for step S4, Select appropriate data fusion methods based on the characteristics and interrelationships of multiple quantitative evaluation values; data fusion methods include weighted average method, principal component analysis (PCA), neural network fusion method, etc. The weighted average method assigns a corresponding weight to each ultrasonic feature influencing factor according to its influence on the grain size, and then performs weighted averaging on multiple quantitative evaluation values to obtain the grain size characterization vector; Principal component analysis (PCA) treats multiple quantitative evaluation values as points in a multidimensional space and uses it to identify the main direction of change of these points, i.e., the principal components. The multiple quantitative evaluation values are then projected onto the principal components to construct a grain size characterization vector. PCA can reduce the dimensionality of the data while retaining the main information of the data. The neural network fusion method constructs a neural network model that takes multiple quantitative evaluation values as input and a grain size representation vector as output. By training the neural network model, it can automatically learn the complex relationship between multiple quantitative evaluation values and output an accurate grain size representation vector. Before data fusion, multiple quantitative evaluation values are preprocessed, including normalization and standardization. Normalization can convert quantitative evaluation values of different dimensions into the same dimension range to avoid the impact of different dimensions on the data fusion results. Standardization can make the quantitative evaluation values have zero mean and unit variance, improving the stability and accuracy of data fusion. According to the selected data fusion method, the multiple pre-processed quantitative evaluation values are fused to obtain the grain size characterization vector.

[0028] In this embodiment, different ultrasonic feature influencing factors reflect the information of grain size from different angles. Through the data fusion method, multiple quantitative evaluation values are integrated, which can comprehensively utilize the advantages of each feature, characterize the grain size more comprehensively and accurately, reduce the deviation that may be caused by a single feature, and thus improve the accuracy of grain size characterization; in actual detection, each quantitative evaluation value may have a certain error; the data fusion process can complement and correct multiple quantitative evaluation values to each other, reduce the influence of single feature error on grain size characterization, and make the final grain size characterization vector closer to the actual situation; before data fusion, multiple quantitative evaluation values are subjected to preprocessing operations such as normalization and standardization; normalization converts quantitative evaluation values of different dimensions into the same dimension range, avoiding interference with data fusion results due to different dimensions; standardization makes the quantitative evaluation value have zero mean and unit variance, improves the stability and comparability of the data, provides high-quality input data for data fusion, and thus enhances the stability and reliability of the data fusion process; data fusion methods such as neural network fusion method can automatically learn the complex nonlinearity between multiple quantitative evaluation values. Relationship, for some characteristic relationships that are difficult to describe with simple mathematical models, it can also be effectively fused; this enables the data fusion process to better adapt to the complex data relationships in actual detection, and improves the reliability of grain size characterization; the principal component analysis method can regard multiple quantitative evaluation values as points in multidimensional space, find the main change direction, and project the quantitative evaluation values onto the principal component to construct a grain size characterization vector; this method can retain the main information of the data while reducing the data dimension, reduce the complexity of data processing, improve calculation efficiency, and avoid problems such as overfitting caused by excessively high data dimensions; the data fusion process integrates multiple quantitative evaluation values into a grain size characterization vector, realizing information concentration and integration; this enables the subsequent grain size inversion model to process and analyze data more conveniently, reducing the difficulty and workload of data processing; according to the characteristics and mutual relationships of multiple quantitative evaluation values, a suitable data fusion method can be selected; different data fusion methods have different characteristics and applicable scopes, so that this step can adapt to the grain size detection needs of different metal materials and under different detection conditions, and improve the versatility of the detection method.

[0029] In some embodiments of the present invention, for step S5, Collect a large number of metal material samples with known grain sizes and obtain the grain size characterization vector corresponding to each sample; Select an appropriate model algorithm based on the relationship between the grain size representation vector and the actual grain size; the model algorithm includes linear regression, support vector machine, and artificial neural network. Linear regression is suitable for situations where there is a linear relationship between the feature and the target variable; support vector machine can handle nonlinear relationships and has good generalization ability in high-dimensional space; artificial neural network has strong nonlinear fitting ability and can learn complex mapping relationships; Use the training set data to train the selected model algorithm, and adjust the model parameters so that the model can accurately fit the training data. During the training process, methods such as cross-validation can be used to evaluate the performance of the model to avoid overfitting. Based on the evaluation results, the model is optimized until the performance meets the requirements. Before inputting the grain size characterization vector of the metal material to be measured into the inversion model, the vector is subjected to the same preprocessing operations as the training data, such as normalization and standardization, to ensure that the format and range of the input data are consistent with the training data; The preprocessed grain size characterization vector is input into the pre-built grain size inversion model. The model will output the grain size prediction result of the metal material to be tested based on the learned mapping relationship.

[0030] In this embodiment, there is a linear or nonlinear relationship between the grain size characterization vector of the metal material and the actual grain size; this step can accurately capture these complex relationships by selecting appropriate model algorithms, such as linear regression to process linear relationships, support vector machines and artificial neural networks to process nonlinear relationships, thereby accurately mapping the grain size characterization vector to the actual grain size value, greatly improving the accuracy of grain size prediction; a large number of metal material samples with known grain sizes are collected and the corresponding grain size characterization vectors are obtained, and these rich data are used for model training; so that the model can learn the general rules between the grain size characterization vector and the actual grain size under different materials and different conditions, avoiding the factors The prediction deviation caused by insufficient data is avoided, ensuring the reliability of the prediction results; once the grain size inversion model is constructed, the grain size characterization vector of the metal material to be tested is input into the model, and the model can quickly output the grain size prediction result according to the learned mapping relationship; so that in the process of industrial production and service, the grain size information of the metal material can be obtained in time to meet the needs of real-time detection and improve production efficiency; the entire prediction process, from data preprocessing to model output results, can be completed automatically by computer programs, without the need for manual complex calculations and judgments; it not only reduces the time and cost of manual operation, but also avoids the influence of human factors on the test results, and improves the degree of automation and stability of the detection.

[0031] In some schemes, multiple embodiments of the present application can be combined and the combined scheme can be implemented. Optionally, some operations in the process of each method embodiment are optionally combined, and / or the order of some operations is optionally changed. In addition, the execution order between the steps of each process is only exemplary and does not constitute a limitation on the execution order between the steps. There can also be other execution orders between the steps. It is not intended to indicate that the execution order is the only order in which these operations can be performed. Ordinary technicians in this field will think of many ways to reorder the operations described herein. In addition, it should be noted that the process details involved in a certain embodiment of this article are also applicable to other embodiments in a similar manner, or different embodiments can be used in combination.

[0032] Furthermore, some steps in the method embodiments may be equivalently replaced with other possible steps. Alternatively, some steps in the method embodiments may be optional and may be deleted in certain usage scenarios. Alternatively, other possible steps may be added to the method embodiments. Furthermore, the various method embodiments may be implemented separately or in combination.

[0033] The basic principles, main features, and advantages of the present invention are shown and described above. Those skilled in the art should understand that the present invention is not limited to the foregoing embodiments. The foregoing embodiments and descriptions are merely illustrative of the principles of the present invention. Various changes and modifications may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and modifications are intended to fall within the scope of the present invention. The scope of protection claimed in the present invention is defined by the appended claims and their equivalents.

Claims

1. A laser ultrasonic prediction method for steel plate grain size, characterized in that: include: Obtain the laser ultrasonic grain size detection process, perform signal feature recognition and extraction on it, and obtain multiple ultrasonic feature influencing factors; For each of the ultrasonic characteristic influencing factors, collecting laser ultrasonic signal data from multiple experiments; Based on a pre-established feature quantification mechanism, feature evaluation is performed on all laser ultrasonic signal data corresponding to the ultrasonic feature influencing factor to obtain a quantitative evaluation value of the ultrasonic feature influencing factor; Performing data fusion on a plurality of the quantitative evaluation values to construct a grain size characterization vector; The grain size characterization vector is input into a pre-built grain size inversion model to obtain a grain size prediction result of the metal material to be tested.

2. The laser ultrasonic prediction method for steel plate grain size according to claim 1, characterized in that: The laser ultrasonic signal data includes time domain signal data, frequency domain signal data and spatial distribution signal data.

3. The laser ultrasonic prediction method for steel plate grain size according to claim 2, characterized in that: The key time domain parameters of the time domain signal data include waveform, amplitude and arrival time.

4. The laser ultrasonic prediction method for steel plate grain size according to claim 2, characterized in that: The key frequency domain parameters of the frequency domain signal data include:

5. The laser ultrasonic prediction method for steel plate grain size according to claim 2, characterized in that: Key parameters of the spatially distributed signal data include signal intensity distribution and phase information.

6. The laser ultrasonic prediction method for steel plate grain size according to claim 1, characterized in that: The factors affecting the construction of the characteristic quantization mechanism include metal type, alloy composition, ultrasonic signal type, ultrasonic signal characteristic stability, laser parameters and environmental factors.

7. The laser ultrasonic prediction method for steel plate grain size according to claim 1, characterized in that: Based on a pre-established feature quantification mechanism, feature evaluation is performed on all signal data corresponding to the ultrasound feature influencing factor to obtain a quantitative evaluation value of the ultrasound feature influencing factor, including: For each ultrasound feature influencing factor, a corresponding feature quantification mechanism is pre-built; Preprocessing the collected laser ultrasonic signal data; According to the pre-built feature quantification mechanism, all signal data corresponding to each ultrasonic feature influencing factor are calculated to obtain the quantitative evaluation value of the feature influencing factor; Verify the calculated quantitative evaluation values.

8. The laser ultrasonic prediction method for steel plate grain size according to claim 1, characterized in that: The grain size characterization vector is input into a pre-built grain size inversion model to obtain the grain size prediction result of the metal material to be tested: Collect metal material samples with known grain sizes and obtain the grain size characterization vector corresponding to each sample; Selecting a model algorithm based on the relationship between the grain size characterization vector and the actual grain size; the model algorithm includes linear regression, support vector machine and artificial neural network; Use the training set data to train the selected model algorithm and adjust the model parameters to make the model accurately fit the training data; During the training process, the cross-validation method is used to evaluate the performance of the model, and the model is optimized based on the evaluation results; Performing preprocessing operations on the grain size characterization vector, including normalization and standardization; The preprocessed grain size characterization vector is input into the pre-built grain size inversion model to output the grain size prediction result of the metal material to be tested.

9. An electronic device for laser ultrasonic prediction of steel plate grain size, comprising a bus, a transceiver, a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the transceiver, the memory, and the processor are connected via the bus, and characterized in that: When the computer program is executed by the processor, the steps of the method according to any one of claims 1 to 8 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.

Images