Metal material rapid annealing effect detection method based on acoustic response

Abstract

The invention relates to the technical field of metal material heat treatment quality detection, in particular to a metal material rapid annealing effect detection method based on acoustic response, which comprises the following steps: step 1, acoustic excitation and synchronous acquisition; step 2, multi-physics field coupling compensation: calculating the phase compensation amount of the vibration signal based on the temperature distribution data and the acoustic characteristic temperature dependence of the material; the method comprises the following steps: acquiring an environment vibration frequency spectrum through a production line mechanical vibration monitoring device, and separating an excitation response component from an original vibration signal by adopting adaptive filtering; 3, annealing sensitive feature extraction; 4, dynamic quality mapping and control: inputting the real-time feature vector into a pre-trained machine learning model, and outputting a recrystallization proportion and a grain size; and when the output parameter exceeds the process threshold value, the control parameter adjustment of the annealing furnace is triggered. Through a non-destructive and accurate acoustic monitoring technology, the annealing effect of the metal material can be efficiently and accurately monitored in a high-speed, high-noise and strong-interference industrial environment.

Description

Technical Field

[0001] The present invention relates to the technical field of metal material heat treatment quality detection technology, and in particular to a metal material rapid annealing effect detection method based on acoustic response. Background Art

[0002] Annealing effect testing of metal materials is a core step in heat treatment quality control, and its accuracy directly affects the mechanical properties, formability, and service reliability of the material. In continuous annealing production lines (such as strip steel and copper wire rolling lines), it is necessary to capture microstructural evolution (such as recrystallization ratio and grain size) in real time under industrial scenarios with high speed operation (line speed ≥ 3m / s), strong vibration interference, and temperature gradients (±50°C). However, existing detection technologies have serious limitations: 1. Technical bottlenecks of existing detection methods: (1) Destructive metallographic inspection: Samples are cut, polished, and etched, and then the grain size is observed under a microscope. This method requires at least 24 hours to complete a single inspection and cannot be used for real-time production line control. The sampling coverage rate is less than 1%, resulting in an inability to reflect the uniformity of the entire strip length, and destructive sampling results in material waste.

[0003] (2) Indirect physical quantity detection method: The electrical resistance method relies on the correlation between resistance and structural defects, but is significantly affected by contact pressure fluctuations and temperature drift. For example, during copper wire annealing testing, probe pressure must be adjusted multiple times to obtain the "lowest resistance value." This results in measurement errors exceeding 15%, and the method cannot distinguish between recrystallization and changes in dislocation density.

[0004] Hardness method: Surface hardness cannot represent deep tissue gradients. Pipe testing requires sectioning and point-by-point testing, taking over 45 minutes per piece. Temperature fluctuations of ±50°C can cause hardness values ​​to drift by up to 10%.

[0005] (3) High-end physical field analysis method X-ray diffraction: The equipment is expensive and requires a vacuum environment. The detection speed is less than 0.5m / s, which cannot be matched with high-speed production lines. Ultrasonic phased array: In annealed coarse-grained materials, the acoustic attenuation is severe, the signal-to-noise ratio is less than 6dB, and the defect detection rate is less than 60%.

[0006] Therefore, there is an urgent need for a method for detecting the rapid annealing effect of metal materials based on acoustic response to solve the above problems. Summary of the Invention

[0007] Based on the above objectives, the present invention provides a method for detecting the rapid annealing effect of metal materials based on acoustic response, comprising: Step 1: Acoustic excitation and synchronous acquisition: A broadband linear frequency modulated acoustic wave signal is emitted to the moving metal strip through a piezoelectric ceramic transducer array, and its frequency band covers the resonance mode in the thickness direction of the material; During the excitation, a non-contact vibration sensor is used to collect the strip surface vibration response signal, and an infrared temperature measuring device is used to obtain the surface temperature distribution; Step 2: Multiphysics coupling compensation: Calculate the phase compensation of the vibration signal based on the temperature distribution data and the temperature dependence of the material's acoustic properties; The environmental vibration spectrum is obtained through the production line mechanical vibration monitoring device, and the excitation response component is separated from the original vibration signal using adaptive filtering; Step 3: Annealing sensitive feature extraction: Perform time-frequency analysis on the compensated signal and select the sub-band with the largest energy concentration change based on the Brillouin scattering characteristics of the metal lattice; Extracting energy proportion parameters and signal decay time constants of the sub-frequency bands; Step 4: Dynamic Quality Mapping and Control: The real-time feature vector is fed into a pre-trained machine learning model to output the recrystallization ratio and grain size. When the output parameter exceeds the process threshold, the annealing furnace control parameter adjustment is triggered.

[0008] Preferably, the method for determining the frequency band in step 1 includes: Pre-establish a library of mapping relationships between material thickness and resonant frequency: Prepare samples of the same material with varying thickness gradients, perform a sweep frequency excitation experiment on each sample, and record the distribution range of its resonant frequency peak value; According to the actual thickness of the strip detected online, the corresponding frequency lower limit and upper limit are matched from the mapping relationship library; The sweep range of the linear frequency modulation signal is set to the matched frequency interval, and the sweep period is dynamically adjusted according to the strip movement speed to ensure that each detection area is fully excited.

[0009] Preferably, the calculation of the phase compensation amount in step 2 includes: The temperature impact sections are divided along the sound wave propagation path, and the length of each section is determined according to the spatial density of infrared temperature measurement points; Obtain the average temperature measurement value of each section, and calculate the sound wave propagation time offset of the section based on the calibration curve of the material sound velocity changing with temperature; Accumulate the propagation time offset of each section and convert it into the phase compensation of the vibration signal; The calibration curve is obtained by laboratory calibration: measuring the sound wave propagation velocity of the same material at different temperature points in a controllable temperature environment, and fitting the sound velocity-temperature function relationship.

[0010] Preferably, the implementation of the adaptive filtering includes: The vibration spectrum collected by the mechanical vibration monitoring device is used as the reference noise signal; Construct a finite impulse response filter and dynamically adjust the filter coefficients through an iterative algorithm to maximize the similarity between the filter output signal and the reference noise signal; Subtract the filter output signal from the original vibration signal to obtain the pure acoustic response component; The termination condition of the iterative algorithm is that the power spectrum density of the residual signal drops below a set ratio.

[0011] Preferably, the sub-band selection method includes: Perform wavelet packet decomposition on the compensated signal to obtain a sub-band set covering the entire frequency band; Calculate the difference in energy concentration between the fully annealed state and the unannealed state for each sub-band; Selecting a continuous sub-band combination with the largest energy concentration difference as a characteristic band, wherein the total bandwidth thereof does not exceed a preset proportion of the full bandwidth; The energy concentration is defined as the ratio of the sub-band signal energy to the full-band signal energy.

[0012] Preferably, the extraction of the decay time constant includes: Extract the envelope of the characteristic sub-band signal and identify the data segment in the signal attenuation stage; Selecting data points above the environmental noise threshold in the attenuation data segment; The selected data points are fitted into an exponential decay curve using a nonlinear fitting algorithm; Calculates the time required for the amplitude to decay to a specific fraction of its initial value based on the fitted curve.

[0013] Preferably, the training method of the machine learning model includes: Prepare a sample set covering different annealing degrees, and simultaneously collect acoustic feature vectors and metallographic detection data for each sample; A regression algorithm with adjustable kernel function is used for training, and the kernel function type is adaptively selected according to the number of samples and feature dimensions; The model hyperparameters are optimized through cross-validation until the mean absolute error between the predicted value and the metallographic detection value is lower than the preset requirement.

[0014] Preferably, the preparation of the sample set includes: Apply gradient annealing process to the same batch of materials to make the recrystallization ratio cover the range of 10% to 100%; The number of samples in each annealing state is determined according to the uniformity of the material structure. The number of samples is increased for materials with poor uniformity. When collecting acoustic features, the production line environment is simulated, and background vibration noise of the same magnitude as that of online detection is introduced.

[0015] Preferably, the triggering of the annealing furnace control parameter adjustment includes: When the recrystallization ratio is lower than the target value, a temperature increase control instruction is generated, and the temperature increase is proportional to the difference between the target value and the measured value; When the grain size exceeds the upper limit, a temperature reduction control instruction is generated, and the temperature drop is proportional to the difference between the measured value and the upper limit; The proportional coefficient is determined by process experiments: the parameter change rate is measured under a fixed temperature adjustment amount, and the average response coefficient of multiple experiments is taken.

[0016] Preferably, the method further includes a model online updating step: Periodically take strip samples for metallographic inspection, and the inspection cycle is set according to the stability of the material batch; Add the acoustic feature vectors and metallographic data of the newly added samples to the training set; When the number of new samples reaches the set proportion of the original data set, the machine learning model is retrained; After the old and new models run in parallel for a predetermined period, the model with the smaller prediction error is selected as the current model.

[0017] Beneficial effects of the present invention: 1. This invention overcomes the destructive nature of metallographic inspection through a real-time monitoring method based on acoustic response. By using a piezoelectric ceramic transducer array and a non-contact vibration sensor, the acoustic response signal of the metal strip can be captured in real time while it is in high-speed motion. The temperature distribution can also be measured simultaneously, enabling comprehensive monitoring of the microstructural evolution during the annealing process. This method eliminates the need for sample collection, avoiding material waste, and provides efficient and accurate real-time data through continuous online monitoring, significantly improving detection efficiency and quality control response speed.

[0018] 2. The detection method of the present invention combines acoustic excitation with a non-contact vibration sensor, solving the problems of contact interference and temperature fluctuations. By leveraging the relationship between the acoustic signal and the thermal properties of the material, the present invention can accurately capture the microstructural changes of the material during the annealing process. Furthermore, through temperature compensation and adaptive filtering techniques, the influence of environmental interference can be effectively eliminated, ensuring the accuracy of the detection signal. This method not only solves the error issues of the resistance method and the hardness method, but also reflects the microstructural state within the material, achieving accurate characterization of deep tissue gradients.

[0019] 3. The present invention solves the shortcomings of high-end physical field analysis methods in high-speed production line applications through a detection method based on acoustic response. Through precise linear frequency modulation acoustic wave excitation signals and efficient signal acquisition technology, the present invention can perform real-time detection under high-speed operation conditions, meeting the detection speed requirements of continuous annealing production lines. Unlike ultrasonic phased arrays, the acoustic response detection of the present invention is not affected by the acoustic attenuation of materials and can effectively avoid the problem of low signal-to-noise ratio. At the same time, by using methods such as time-frequency analysis and attenuation time constant extraction, the microstructural changes of metal materials can be accurately captured, the detection accuracy is greatly improved, and the limitations of traditional technologies are avoided. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0021] Figure 1 is a flow chart of the steps of the method of the present invention; Figure 2 A flow chart showing the steps of triggering the adjustment of the annealing furnace control parameters according to the method of the present invention; Figure 3 This is a flowchart of the steps for online updating of the model in the method of the present invention. DETAILED DESCRIPTION

[0022] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments. It is also noted that, to provide a more detailed description, the following embodiments are best and preferred embodiments, and those skilled in the art may employ alternative methods for implementing certain known technologies. Furthermore, the accompanying drawings are intended only to provide a more detailed description of the embodiments and are not intended to limit the present invention.

[0023] See Figure 1-Figure 3 , an embodiment of the present invention provides a method for detecting the rapid annealing effect of metal materials based on acoustic response. In step 1, first, a piezoelectric ceramic transducer array is used to emit a broadband linear frequency modulation sound wave signal. The frequency band of this signal can cover the resonance mode of the metal strip in the thickness direction, ensuring that various microstructural changes produced by the metal material during the annealing process can be stimulated. After penetrating the metal strip, these sound wave signals will interact with its internal microstructure and change its propagation characteristics. At the same time, the non-contact vibration sensor synchronously collects the vibration response signals on the surface of the strip. These signals reflect the changes in the internal lattice of the metal and can capture the microscopic evolution process of the material in real time. In addition, the surface temperature distribution data of the material is obtained by an infrared temperature measuring device to provide temperature change information for subsequent analysis and ensure the accuracy of the detection process.

[0024] In step 2, this method calculates the phase compensation for the vibration signal based on the temperature distribution data. Temperature significantly affects the acoustic properties of materials, so compensation based on temperature data avoids errors caused by temperature fluctuations. Furthermore, the ambient vibration spectrum is acquired through production line mechanical vibration monitoring. Adaptive filtering techniques are used to separate the noise generated by the mechanical vibration and the excitation signal from the raw vibration signal, ensuring that the extracted signal represents only the actual changes within the material.

[0025] In step 3, after compensation, the vibration signal undergoes time-frequency analysis. Due to the lattice Brillouin scattering characteristics of metal materials, microstructural changes during the annealing process can lead to significant changes in the energy concentration of different frequency components. By analyzing the energy distribution of different sub-bands, the frequency band with the largest change in energy concentration is selected. Based on this, relevant features are extracted, including parameters such as the signal decay time constant and the sub-band energy fraction. These features can effectively reflect the grain evolution and degree of recrystallization of the metal material during the annealing process.

[0026] In step 4, the feature vectors extracted in real time are fed into a pre-trained machine learning model to determine key parameters such as recrystallization ratio and grain size. These parameters directly impact the material's mechanical properties and service reliability. When the real-time monitored parameters exceed the process threshold, a control signal is automatically triggered to adjust the annealing furnace's temperature, heating time, and other control parameters, thereby achieving precise control of the annealing process and ensuring that the material meets the expected quality standards.

[0027] The application of acoustic response technology not only enables real-time, non-destructive detection of the annealing effect of metal strips, but also maintains high precision under complex operating conditions such as high temperature, high speed, and high vibration. Furthermore, the combination of multi-physics field coupling compensation and time-frequency analysis effectively improves the accuracy and robustness of detection. Furthermore, the application of machine learning models enables more intelligent control of the annealing process, enabling dynamic adjustment of process parameters to improve production efficiency and product quality.

[0028] In one possible implementation, to accurately determine the frequency band of the excitation signal, a series of samples of the same metal material with varying thickness gradients are first prepared. A swept-frequency excitation experiment is performed on these samples, and the peak resonant frequency of each sample at different thicknesses is recorded. This process captures the variations in the acoustic properties of the metal material at varying thicknesses and provides basic data for subsequent frequency matching. Through multiple experiments, a mapping relationship between material thickness and resonant frequency is obtained, forming a mapping relationship library.

[0029] During actual production, the thickness of the strip may vary, necessitating real-time monitoring of the strip thickness and matching the corresponding frequency range from an established mapping library. Specifically, the mapping library provides lower and upper frequency limits based on the actual thickness of the strip, ensuring that the frequency range of the excitation signal matches the resonant frequency of the strip, thereby ensuring that the sound waves effectively excite the material's resonant modes.

[0030] Once the frequency range is determined, the next step is to set the linear frequency modulation signal sweep range to ensure that the signal sweep covers the matching frequency band. During the strip's motion, to avoid signal omissions or repeated excitation, the sweep period is dynamically adjusted. This adjustment is based on the strip's actual motion speed, ensuring that each detection area is fully excited and that any potential changes in the annealing effect are not missed.

[0031] By establishing a mapping library between thickness and resonant frequency, the appropriate excitation frequency range can be automatically selected based on the actual thickness of the metal strip, significantly improving the accuracy and specificity of the detection signal. Furthermore, by dynamically adjusting the sweep frequency period, the complete excitation of the signal can be ensured, resulting in more accurate acoustic response data and enabling high-precision annealing effect detection. Compared to traditional fixed-frequency excitation methods, this method can adapt to changes in strip thickness, increasing the flexibility and adaptability of the entire detection process.

[0032] In one possible implementation, the acoustic wave propagation path is divided into several temperature-affecting segments based on the density of spatial temperature points captured by an infrared temperature measurement device. The length of each segment is based on the spacing between the infrared temperature measurement points, ensuring that temperature changes within each segment are accurately reflected. The principle of segmentation is to ensure the spatial resolution of the temperature information while avoiding excessive computational complexity caused by too many segments.

[0033] Within each segment, measurements are collected at corresponding temperature points and the average temperature is calculated. Using a pre-established laboratory sound velocity-temperature calibration curve for the material, the average temperature for that segment is mapped to the sound wave propagation velocity, thereby calculating the sound wave propagation time offset for that segment. This calibration curve is obtained by measuring the sound wave propagation velocity of the same metal material at different temperatures in a temperature-controlled environment and fitting it into a continuous sound velocity-temperature function, achieving a precise mapping of temperature to sound velocity.

[0034] The acoustic wave propagation time offsets calculated for each section along the propagation path are accumulated to obtain a total propagation time offset. This time offset is then converted into a phase compensation value for the vibration signal. This allows the collected vibration signal to eliminate the propagation velocity variation error caused by temperature and accurately reflect the acoustic properties of the material itself.

[0035] The above steps effectively overcome the effects of temperature gradients on acoustic wave propagation during rapid annealing, allowing the vibration signal to more accurately reflect the material's state and improving the accuracy of annealing effect detection. Furthermore, this method, which requires no contact measurement, enables real-time detection on moving strips, adapting to high-speed production line environments. It avoids signal distortion caused by temperature fluctuations in traditional methods, thereby enhancing the reliability and repeatability of detection results.

[0036] In one possible implementation, a mechanical vibration monitoring device is first used to collect vibration signals from the strip during the annealing process. These vibration signals contain external noise components, such as equipment operation or environmental vibrations. The collected vibration spectrum is used as a reference noise signal for subsequent adaptive filtering.

[0037] To remove noise, a finite impulse response (FIR) filter is constructed. This filter uses an iterative algorithm to adjust its coefficients to optimize the filtering effect. With each iteration of the filter, the algorithm dynamically adjusts the filter coefficients based on the characteristics of the reference noise signal, maximizing the similarity between the filter output signal and the reference noise signal. In other words, the filter continuously adjusts its response to maximize noise rejection.

[0038] After processing by the adaptive filter, the filter output signal represents the original vibration signal after noise is removed. Next, the filter output signal is subtracted from the original vibration signal to obtain the pure acoustic response component. This pure acoustic signal reflects the actual acoustic properties of the metal material and can be used to determine the annealing effect.

[0039] The adaptive filtering iteration process stops when a specific condition is met. Specifically, the algorithm stops when the power spectral density of the residual signal (i.e., the energy of the noise component) falls below a set percentage. This condition ensures the accuracy of the filtering process, ensuring that the final output signal no longer contains significant noise components.

[0040] Adaptive filtering effectively separates noise from pure acoustic response components, significantly improving signal clarity and accuracy. This approach accurately extracts acoustic signals reflecting the annealing effects of materials, even in the high-noise environment of the annealing process. Furthermore, the dynamic adjustment of the iterative algorithm makes the filtering process more intelligent and flexible, adapting to varying noise conditions and maintaining efficient and stable detection performance in changing production environments.

[0041] In one possible implementation, the phase-compensated signal is first subjected to wavelet packet decomposition. Wavelet packet decomposition is a multi-scale signal analysis method that decomposes the original signal into a set of subbands within different frequency ranges. Each subband represents the energy information of a specific frequency range in the original signal, thereby clarifying the signal's frequency domain characteristics. This decomposition covers all subbands of the entire frequency band, providing comprehensive frequency domain information for subsequent analysis.

[0042] After completing the wavelet packet decomposition, the difference in energy concentration between the fully annealed and unannealed states is calculated for each subband. Energy concentration is the ratio of the signal energy in a particular subband to the full-band signal energy, reflecting the relative concentration of signal energy within that frequency band. By comparing the energy concentration of each subband under different annealing conditions, the sensitivity of that subband to the annealing effect can be assessed. Subbands with large energy concentration differences generally better reflect changes in the material's annealing state.

[0043] After calculating the energy concentration differences across all subbands, the combination of consecutive subbands with the largest energy concentration difference is selected as the characteristic band. This characteristic band combination, as the signal portion that best reflects the annealing effect of the metal material, will be used in subsequent annealing effect analysis. To avoid over-segmentation, the total bandwidth of the selected characteristic bands does not exceed a preset proportion of the full bandwidth. This ensures simplicity in analysis while avoiding noise interference caused by excessive bandwidth.

[0044] By employing wavelet packet decomposition and energy concentration difference analysis, the frequency domain characteristics associated with the annealing state can be accurately captured. This method can identify frequency bands that exhibit significant differences between annealing states. These frequency bands are most sensitive to the annealing process, effectively improving detection accuracy. Furthermore, the selected characteristic frequency bands are highly indicative of annealing effects, effectively removing noise interference and reducing the influence of irrelevant signals, ensuring the reliability and accuracy of the detection results.

[0045] In one possible implementation, envelope extraction is first performed on the selected characteristic sub-band signal. The envelope is the outer contour curve of the signal, which reflects the amplitude variation trend of the signal. During the annealing process, the acoustic response signal of the metal material gradually decays over time. The envelope can clearly depict this decay process and help identify the decay stage of the signal.

[0046] By observing the envelope, we can identify the decay phases of the signal. These decay segments typically start with a maximum amplitude and gradually decrease. The decay process is an important indicator for studying the effects of material annealing, as the material's physical properties change during the annealing process, causing the decay pattern of the acoustic response signal to change accordingly.

[0047] After identifying the attenuated data segments, we further screen out data points above the ambient noise threshold. Environmental noise can interfere with signal analysis, so selecting only those signal data points that are significantly above the noise level for analysis effectively avoids interference from invalid information and improves the accuracy of the analysis results.

[0048] The effective data points after screening are fitted into an exponential decay curve using a nonlinear fitting algorithm. The exponential decay curve is described by the following mathematical expression: ; in, is the amplitude at time t, is the initial amplitude, is the decay constant, which represents the decay rate of the signal. This decay constant can be accurately solved using a nonlinear fitting algorithm.

[0049] Based on the fitted exponential decay curve, calculate the time required for the amplitude to decay to a specified fraction of its initial value. For example, you can set a specific fraction (e.g., decay to 50% of the initial value) and calculate the time required to reach that fraction from the fitted curve. This time is the decay time constant and reflects the rate at which the signal decays.

[0050] This decay time constant extraction method accurately reflects the changing characteristics of the acoustic response of metal materials during rapid annealing. The decay time constant provides a quantitative indicator that clearly describes the annealing effect of metal materials. Compared with traditional methods, this method can more accurately analyze the signal decay process, avoid interference from environmental noise, and improve detection accuracy. In addition, the application of a nonlinear fitting algorithm makes the decay time calculation more precise, providing reliable data support for real-time monitoring and analysis.

[0051] In one possible implementation, a sample set covering different annealing degrees needs to be prepared first. Each sample should be collected simultaneously by recording two pieces of data: acoustic eigenvectors and metallographic data. Acoustic eigenvectors reflect the acoustic response of metal materials under different annealing conditions, while metallographic data provides information about the metal's microstructure. By simultaneously collecting these two types of data, a wealth of information can be provided to the machine learning model, enabling the model to extract features related to changes in the material's microstructure from the acoustic signal, thereby more accurately predicting the annealing effect.

[0052] Next, a regression algorithm with an adjustable kernel function is trained on the sample set. Regression algorithms are a typical supervised learning method that makes predictions by learning the mapping relationship between input features and output results between samples. Here, the input features are acoustic feature vectors, and the output is metallographic inspection data. The regression model uses kernel function technology, which effectively maps data into a high-dimensional space and improves the model's ability to fit complex data relationships. The kernel function type is selected dynamically, adaptively adjusting based on the number of samples and feature dimensionality. This means that for different sample sets, the model will select the most appropriate kernel function to ensure the best fit.

[0053] To optimize model performance, cross-validation is used for model training and validation. Cross-validation involves dividing a sample set into multiple subsets, alternating between using one subset as the validation set and the remaining subset as the training set. Cross-validation reduces overfitting of the model to the training data and improves its generalization ability. Furthermore, during training, model hyperparameters (such as kernel function parameters and regularization coefficients) are continuously adjusted through cross-validation until an optimal combination of hyperparameters is found. The goal of this step is to minimize the model's prediction error.

[0054] The ultimate goal of the training process is to ensure that the mean absolute error (MAE) between the model's predictions and metallographic test results is below a preset requirement. By continuously optimizing the kernel function and hyperparameters, the model is ultimately able to provide accurate predictions on real data. This evaluation criterion ensures the model's predictive power and practicality, ensuring its effectiveness in real-world applications.

[0055] This machine learning training method efficiently establishes a precise mapping between acoustic features and the annealing effects of metal materials. Compared to traditional detection methods, detection based on acoustic responses is not only highly efficient but also allows for sensitive and accurate predictions under different annealing conditions. Through a regression algorithm with an adjustable kernel function, the model can adapt to the characteristics of different sample sets, ensuring the algorithm's ability to fit complex data. At the same time, cross-validation and hyperparameter optimization steps effectively avoid overfitting, improve the model's generalization ability, and ensure its ability to stably and reliably predict the annealing effects of metal materials in practical applications.

[0056] In one possible implementation, first, a gradient annealing process is applied to the same batch of materials, which means that during the annealing process, the material will undergo different degrees of heating and cooling to promote the formation of different recrystallization ratios. Specifically, the annealing process will be controlled within a range so that the recrystallization ratio gradually changes from 10% to 100%. This process can simulate the microstructural changes of materials under different annealing states, providing rich samples for subsequent acoustic characteristics and metallographic testing. These different annealing states can cover the entire process from slight annealing to complete recrystallization, ensuring that the model can learn the differences in acoustic responses under different annealing degrees.

[0057] During sample preparation, the number of samples per annealing state is adjusted based on the material's microstructure uniformity. Materials with poor microstructure uniformity often have more complex and uneven microstructures. These variations can affect the extraction of acoustic features. Therefore, to better capture these microscopic differences, the number of samples of such materials needs to be increased. Increasing the number of samples, especially for materials with heterogeneous microstructures, makes the training set more comprehensive, reduces bias caused by insufficient samples, and thus improves the model's ability to learn from different material states.

[0058] To ensure the data's practical application value, the acoustic feature collection process simulates the production line environment. This means introducing background vibration noise equivalent to that encountered during online testing. This design ensures that the acoustic signals in the training set are more closely aligned with real-world application scenarios, as online testing is often accompanied by varying degrees of ambient noise. By incorporating background noise during sample collection, the model learns how to extract effective features from noisy signals during training, thereby improving its interference tolerance and reliability in real-world environments.

[0059] This sample set preparation method ensures the representativeness and diversity of the model training data. The gradient annealing process covers the entire annealing process from low recrystallization to high recrystallization, allowing the model to capture acoustic changes under different annealing degrees. Adjusting the number of samples according to the uniformity of the material structure can ensure that enough samples are collected in materials with uneven structures, thereby improving the model's adaptability to complex structures. Simulating the production line environment to collect acoustic features and adding background noise enhances the model's adaptability to the actual production environment and can effectively improve the model's accuracy and robustness in practical applications. In short, this method improves the quality of the data, enhances the model's predictive ability, and ensures the feasibility and practical application value of the detection method.

[0060] In one possible implementation, first, the recrystallization ratio of the material is monitored in real time. When the recrystallization ratio is found to be lower than the set target value, the system generates a temperature increase control instruction, the purpose of which is to accelerate the recrystallization process by increasing the temperature. The amount of temperature increase is proportional to the difference between the target value and the actual measured value. Specifically, the greater the difference between the target value and the measured value, the greater the temperature increase, which can prompt the recrystallization ratio to reach the target requirement as soon as possible. This step precisely controls the recrystallization process during the annealing process by adjusting the rate of temperature change.

[0061] At the same time, the grain size is monitored. If the measurement results indicate that the grain size has exceeded the predetermined upper limit, the system generates a temperature reduction control instruction to prevent further grain growth. The amount of temperature reduction is proportional to the difference between the measured grain size and the upper limit. That is, the larger the grain size, the greater the temperature reduction, effectively preventing further grain growth and ensuring that the grain size remains within the target range.

[0062] To ensure accurate and stable temperature control, selecting the proportionality factor is crucial. This factor is determined through process experiments. During these experiments, the impact of temperature changes on various material parameters (such as recrystallization ratio and grain size) is measured at a fixed temperature adjustment. The average response factor is calculated from the results of multiple experiments to determine the optimal proportionality factor. This proportionality factor ensures that the temperature control responsiveness matches actual production requirements, avoiding excessive or insufficient temperature changes.

[0063] This annealing furnace temperature adjustment method based on proportional control can achieve precise temperature control and ensure dynamic adjustment of the recrystallization ratio and grain size during the annealing process. First, through temperature rise control, the recrystallization ratio can be effectively improved to ensure that the material achieves the expected microstructure; second, through temperature drop control, excessive grain growth can be avoided and the stability of the grain size can be guaranteed. This method not only optimizes the annealing process and improves the quality of metal materials, but also can respond to changes in real time and automatically adjust temperature control, thereby improving production efficiency and material performance. Finally, the proportional coefficient determined by experiment makes temperature adjustment more precise, avoids quality fluctuations caused by over-adjustment, and ensures the reliability and consistency of the annealing effect.

[0064] In one possible implementation, to ensure that the model can adapt to changes in material batches in real time, it is first necessary to regularly take strip samples from the production line for metallographic inspection. Metallographic inspection can provide information on the material's microstructure, such as the recrystallization ratio and grain size, which are crucial for understanding the annealing effect of the material. The inspection cycle is set based on the stability of the material batch. For materials with large batch fluctuations, the inspection cycle will be shortened accordingly; for materials with good batch stability, the inspection cycle can be appropriately extended. This measure ensures the representativeness and timeliness of sampling, and avoids affecting the accuracy of test results due to large fluctuations in material properties.

[0065] After each metallographic inspection, the obtained metallographic data and the corresponding acoustic feature vectors are added to the training set. This data fusion process enables the model to learn from an increasing number of new samples, thereby capturing the patterns and characteristic changes in more samples. This step ensures the real-time update of the training set data and maintains the dynamic adaptability of the model.

[0066] When the number of new samples reaches a set ratio of the original dataset, the system triggers retraining of the machine learning model. This ratio is typically determined based on the changing patterns of data volume and the required accuracy of the model. By introducing new samples, the model can learn more data reflecting the current production status, further improving its predictive power and accuracy for the metal annealing process.

[0067] After the new model completes training, the old and new models are run in parallel for a predetermined period. This parallel running allows comparison of their predictions under the same conditions to assess whether the new model outperforms the old one. This comparison ensures the rationality of the model update and avoids errors caused by issues such as instability or overfitting of the new model.

[0068] After the old and new models have run in parallel, the system will select the model with the smallest prediction error as the current model. This selection process is based on the accuracy of the actual prediction results to ensure that the final selected model can provide the best prediction results and avoid the impact of the annealing effect detection accuracy due to model outdatedness or reduced accuracy.

[0069] Regularly collecting samples for metallographic examination ensures the representativeness and timeliness of test data, avoiding prediction bias caused by data lag. Furthermore, the addition of new samples allows the model to gradually adapt to changes in material batches, improving its adaptability to different batches of materials. A proportionally controlled model update mechanism ensures that the model is always in an optimal state, avoiding overtraining or undertraining. Running the old and new models in parallel effectively reduces the risks associated with model updates, making each update more reliable. This online update mechanism ensures the long-term stability and reliability of the annealing effect detection method.

[0070] The following is a detailed explanation using examples: This example uses the steel annealing process as an example to describe a method for detecting the effectiveness of rapid annealing of metal materials based on acoustic response. This method combines metallographic inspection data with acoustic signatures, using a machine learning model to update and adjust the annealing process in real time, ensuring that the material's annealing effect meets predetermined standards.

[0071] In the embodiment of the present invention, the metal material type is low-alloy high-strength steel (such as Q345 steel), which is commonly used for high-strength components in the fields of construction, bridges, etc.

[0072] Annealing process: Annealing is carried out in a medium frequency electric furnace. When the steel strip passes through the annealing furnace, the temperature is controlled to rise from 300°C to 900°C, and then gradually cool to room temperature.

[0073] Sensors are installed to collect acoustic response data as the steel strip passes through the annealing furnace. The acoustic signals collected by the sensors are converted into spectral data using Fourier transform, resulting in a frequency response signature. The acquisition frequency range is set to 20 Hz to 5 kHz, with sampling once per second for 10 seconds.

[0074] The acoustic signal is converted into a frequency domain signal using Fast Fourier Transform (FFT) and the following features are extracted: Peak Frequency: represents the frequency with the highest energy in the acoustic response.

[0075] Total Power: The power spectrum area of ​​the signal, which represents the total energy of the signal.

[0076] Spectral shape characteristics: The state of the material is further analyzed by calculating the morphological characteristics of the spectrum (such as average amplitude, bandwidth, etc.).

[0077] Every 500 meters of the production line, a strip sample is cut for metallographic examination to measure the grain size (D) and recrystallization ratio (R). Metallographic examination is carried out using standard metallographic sectioning methods using an optical microscope or scanning electron microscope (SEM).

[0078] Assume that there are currently 1,000 samples, each of which contains acoustic characteristics (such as peak frequency, total power, spectral shape characteristics, etc.) and metallographic data (such as grain size, recrystallization ratio).

[0079] The training set data includes pairs of acoustic features and metallographic data, and the target variables are grain size (D) and recrystallization ratio (R).

[0080] A support vector machine (SVM) regression model is selected for training. The goal of the model is to predict metallographic data based on acoustic features. The model training steps are as follows: The acoustic features of the training set are normalized so that the mean of the data is 0 and the standard deviation is 1.

[0081] Using the scikit-learn library in Python, the support vector machine regression model is trained using the RBF kernel function. The training parameters are as follows: C (regularization parameter): 10 γ (kernel function parameter): 0.01 Epsilon (tolerance): 0.1 After every 1000 meters of strip is produced, samples are taken for metallographic examination to measure their grain size (D) and recrystallization ratio (R).

[0082] Add the newly added metallographic data and acoustic features (collected in real time by acoustic sensors) to the training set. Assume that 10 new samples are added each time.

[0083] When the number of new samples reaches 10% of the original training set (that is, 100 new samples), retraining of the machine learning model is triggered.

[0084] When new data is added, the old model and the new model are run in parallel, their prediction errors on the new test set are compared, and the model with the smaller error is selected as the current model.

[0085] The mean square error (MSE) is chosen as the evaluation criterion.

[0086] MSE of the old model: 0.022 MSE of the new model: 0.018 Based on the comparison results, the new model was selected for continued use because it had a smaller prediction error.

[0087] Based on the metallographic inspection results and the acoustic response data, the temperature of the annealing furnace can be adjusted in real time to ensure that the material reaches the expected grain size and recrystallization ratio. The temperature adjustment formula is as follows: ; in: is the temperature value that needs to be adjusted (unit: ℃).

[0088] and is the target grain size and recrystallization ratio.

[0089] and The current metallographic data.

[0090] =1.5 and =2.0 is the adjustment coefficient, which is determined based on experimental data.

[0091] This method was used for annealing during a batch production process, and metallographic examination was performed to compare the annealing effects of this method with those of traditional annealing methods.

[0092] Traditional annealing method: without real-time monitoring of acoustic response, relying on fixed temperature and time settings, resulting in poor grain size uniformity and large fluctuations in recrystallization ratio.

[0093] The method of the present invention adjusts the temperature of the annealing furnace in real time, and adjusts the annealing process in real time according to acoustic characteristics and metallographic data, thereby ensuring a more uniform annealing effect.

[0094] Traditional method: The standard deviation of grain size uniformity is 12.5μm, and the fluctuation of recrystallization ratio is 7%.

[0095] The method of the present invention has a standard deviation of grain size uniformity of 5.8 μm and a fluctuation of recrystallization ratio of 3%.

[0096] By comparison, it can be seen that the method of the present invention can significantly improve the uniformity of the annealing process, reduce the fluctuation of the grain size, and thus improve the mechanical properties and reliability of the steel.

[0097] This method combines acoustic response with metallographic data, along with an online update mechanism based on machine learning, to monitor and adjust the annealing effect of metal materials in real time. Comparative experiments have shown that this method can effectively improve the control accuracy of the annealing process, reduce fluctuations in material properties during the production process, and enhance the quality and stability of the final product.

[0098] The present invention encompasses any alternatives, modifications, equivalents, and solutions that fall within the spirit and scope of the present invention. To provide a thorough understanding of the present invention, specific details are described in detail below in connection with the preferred embodiments of the present invention, but those skilled in the art will be able to fully understand the present invention without these detailed descriptions. Furthermore, to avoid unnecessary confusion regarding the essence of the present invention, well-known methods, processes, procedures, components, and circuits have not been described in detail.

[0099] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. A method for detecting the rapid annealing effect of metal materials based on acoustic response, characterized in that: include: Step 1: Acoustic excitation and synchronous acquisition: A broadband linear frequency modulated acoustic wave signal is emitted to the moving metal strip through a piezoelectric ceramic transducer array, and its frequency band covers the resonance mode in the thickness direction of the material; During the excitation, a non-contact vibration sensor is used to collect the strip surface vibration response signal, and an infrared temperature measuring device is used to obtain the surface temperature distribution; Step 2: Multiphysics coupling compensation: Calculate the phase compensation of the vibration signal based on the temperature distribution data and the temperature dependence of the material's acoustic properties; The environmental vibration spectrum is obtained through the production line mechanical vibration monitoring device, and the excitation response component is separated from the original vibration signal using adaptive filtering; Step 3: Annealing sensitive feature extraction: Perform time-frequency analysis on the compensated signal and select the sub-band with the largest energy concentration change based on the Brillouin scattering characteristics of the metal lattice; Extracting energy proportion parameters and signal decay time constants of the sub-frequency bands; Step 4: Dynamic Quality Mapping and Control: The real-time feature vector is fed into a pre-trained machine learning model to output the recrystallization ratio and grain size. When the output parameter exceeds the process threshold, the annealing furnace control parameter adjustment is triggered.

2. The method for detecting the rapid annealing effect of metal materials based on acoustic response according to claim 1, characterized in that: The method for determining the frequency band in step 1 includes: Pre-establish a library of mapping relationships between material thickness and resonant frequency: Prepare samples of the same material with varying thickness gradients, perform a sweep frequency excitation experiment on each sample, and record the distribution range of its resonant frequency peak value; According to the actual thickness of the strip detected online, the corresponding frequency lower limit and upper limit are matched from the mapping relationship library; The sweep range of the linear frequency modulation signal is set to the matched frequency interval, and the sweep period is dynamically adjusted according to the strip movement speed to ensure that each detection area is fully excited.

3. The method for detecting the rapid annealing effect of metal materials based on acoustic response according to claim 1, characterized in that: The calculation of the phase compensation amount in step 2 includes: The temperature impact sections are divided along the sound wave propagation path, and the length of each section is determined according to the spatial density of infrared temperature measurement points; Obtain the average temperature measurement value of each section, and calculate the sound wave propagation time offset of the section based on the calibration curve of the material sound velocity changing with temperature; Accumulate the propagation time offset of each section and convert it into the phase compensation of the vibration signal; The calibration curve is obtained by laboratory calibration: measuring the sound wave propagation velocity of the same material at different temperature points in a controllable temperature environment, and fitting the sound velocity-temperature function relationship.

4. The method for detecting the rapid annealing effect of metal materials based on acoustic response according to claim 1, characterized in that: The implementation of the adaptive filtering includes: The vibration spectrum collected by the mechanical vibration monitoring device is used as the reference noise signal; Construct a finite impulse response filter and dynamically adjust the filter coefficients through an iterative algorithm to maximize the similarity between the filter output signal and the reference noise signal; Subtract the filter output signal from the original vibration signal to obtain the pure acoustic response component; The termination condition of the iterative algorithm is that the power spectrum density of the residual signal drops below a set ratio.

5. The method for detecting the rapid annealing effect of metal materials based on acoustic response according to claim 1, characterized in that: The method for selecting the sub-frequency band includes: Perform wavelet packet decomposition on the compensated signal to obtain a sub-band set covering the entire frequency band; Calculate the difference in energy concentration between the fully annealed state and the unannealed state for each sub-band; Selecting a continuous sub-band combination with the largest energy concentration difference as a characteristic band, wherein the total bandwidth thereof does not exceed a preset proportion of the full bandwidth; The energy concentration is defined as the ratio of the sub-band signal energy to the full-band signal energy.

6. The method for detecting the rapid annealing effect of metal materials based on acoustic response according to claim 1, characterized in that: The extraction of the decay time constant includes: Extract the envelope of the characteristic sub-band signal and identify the data segment in the signal attenuation stage; Selecting data points above the environmental noise threshold in the attenuation data segment; The selected data points are fitted into an exponential decay curve using a nonlinear fitting algorithm; Calculates the time required for the amplitude to decay to a specific fraction of its initial value based on the fitted curve.

7. The method for detecting the rapid annealing effect of metal materials based on acoustic response according to claim 1, characterized in that: The training method of the machine learning model includes: Prepare a sample set covering different annealing degrees, and simultaneously collect acoustic feature vectors and metallographic detection data for each sample; A regression algorithm with adjustable kernel function is used for training, and the kernel function type is adaptively selected according to the number of samples and feature dimensions; The model hyperparameters are optimized through cross-validation until the mean absolute error between the predicted value and the metallographic detection value is lower than the preset requirement.

8. The method for detecting the rapid annealing effect of metal materials based on acoustic response according to claim 7, characterized in that: The preparation of the sample set includes: Apply gradient annealing process to the same batch of materials to make the recrystallization ratio cover the range of 10% to 100%; The number of samples in each annealing state is determined according to the uniformity of the material structure. The number of samples is increased for materials with poor uniformity. When collecting acoustic features, the production line environment is simulated, and background vibration noise of the same magnitude as that of online detection is introduced.

9. The method for detecting the rapid annealing effect of metal materials based on acoustic response according to claim 1, characterized in that: The triggering annealing furnace control parameter adjustment includes: When the recrystallization ratio is lower than the target value, a temperature increase control instruction is generated, and the temperature increase is proportional to the difference between the target value and the measured value; When the grain size exceeds the upper limit, a temperature reduction control instruction is generated, and the temperature drop is proportional to the difference between the measured value and the upper limit; The proportional coefficient is determined by process experiments: the parameter change rate is measured under a fixed temperature adjustment amount, and the average response coefficient of multiple experiments is taken.

10. The method for detecting the rapid annealing effect of metal materials based on acoustic response according to claim 1, characterized in that: It also includes the model online update step: Periodically take strip samples for metallographic inspection, and the inspection cycle is set according to the stability of the material batch; Add the acoustic feature vectors and metallographic data of the newly added samples to the training set; When the number of new samples reaches the set proportion of the original data set, the machine learning model is retrained; After the old and new models run in parallel for a predetermined period, the model with the smaller prediction error is selected as the current model.

Images