Metal sheet chemical component analysis method and system based on spectral analysis

Abstract

The invention relates to the technical field of metal material detection, in particular to a metal sheet chemical component analysis method and system based on spectral analysis, and solves the technical problem of low measurement precision in the prior art. The method comprises the following steps: respectively acquiring detection data of each acquisition point of the metal sheet during current detection and multiple historical detection; the detection data comprises real-time working condition data and a reference spectrum of the corresponding acquisition point; for each acquisition point, determining at least one similar historical detection for the acquisition point from the multiple historical detections according to the difference degree between the real-time working condition data of the acquisition point during the current detection and the multiple historical detections; for each acquisition point, correcting the reference spectrum of the acquisition point during current detection according to the reference spectrum of the acquisition point during at least one similar historical detection to obtain a corrected spectrum of the acquisition point; and carrying out chemical component analysis on the metal sheet according to the correction spectrum of each collection point.

Description

Technical Field

[0001] This invention relates to the field of metal material testing technology, specifically to a method and system for analyzing the chemical composition of thin metal plates based on spectral analysis. Background Technology

[0002] With the rapid development of intelligent manufacturing technology, thin metal sheets, as a core raw material in fields such as machinery manufacturing, automotive industry, and electronic equipment, have their chemical composition precisely controlled, directly affecting the quality stability, process adaptability, and safety of end products. In thin metal sheet production lines, rapid and accurate chemical composition analysis is a key technological support for real-time process optimization, product quality grading, and intelligent sorting, and is of great significance for improving production efficiency and reducing production costs.

[0003] Currently, among the methods used in the industry for detecting the chemical composition of thin metal sheets, traditional wet chemical analysis is difficult to meet the real-time detection requirements of modern production lines due to its complex operation, long processing time, and destructive nature to samples. The mainstream detection schemes at present mostly employ fixed-point or line-scan spectroscopy techniques, acquiring the spectral signals of the thin metal sheets and combining them with calibration models to invert the chemical composition. However, the production process of thin metal sheets is a dynamic and complex system. Production conditions (such as sheet temperature gradient, sheet warping, surface oxide film thickness, and running tension) fluctuate in real time. These dynamic changes in operating conditions can cause systematic drifts in the spectral signals, such as baseline shifts and characteristic peak distortions. This makes calibration models established based on standard laboratory static conditions prone to failure, ultimately resulting in insufficient measurement accuracy and poor stability in chemical composition analysis, failing to meet the high reliability requirements of intelligent manufacturing for detection data. Summary of the Invention

[0004] To address the technical problem of low measurement accuracy in existing technologies, the present invention aims to provide a method and system for chemical composition analysis of thin metal plates based on spectral analysis. The specific technical solution adopted is as follows: This application provides a method for chemical composition analysis of thin metal plates based on spectral analysis, including: The detection data for each collection point of the metal sheet is acquired during the current detection and multiple historical detections; the detection data includes real-time operating data and reference spectrum of the corresponding collection point. For each collection point, based on the degree of difference between the real-time operating data of the collection point during the current detection and multiple historical detections, at least one similar historical detection is determined for the collection point from the multiple historical detections; For each acquisition point, the reference spectrum of the acquisition point at the current detection time is corrected based on the reference spectrum of the acquisition point at the at least one similar historical detection time to obtain the corrected spectrum of the acquisition point; The chemical composition of the metal sheet is analyzed based on the calibrated spectrum at each collection point.

[0005] This application provides a system for analyzing the chemical composition of thin metal plates based on spectral analysis, including: The data acquisition module is used to acquire the detection data of each collection point of the metal sheet during the current detection and multiple historical detections; the detection data includes the real-time operating condition data and reference spectrum of the corresponding collection point; The historical matching module is used to determine at least one similar historical detection for each collection point based on the degree of matching between the real-time operating data of the collection point during the current detection and multiple historical detections. A spectral correction module is used to correct the reference spectrum of the acquisition point at the current detection time based on the reference spectrum of the acquisition point at the at least one similar historical detection time for each acquisition point, so as to obtain the corrected spectrum of the acquisition point. The component analysis module is used to perform chemical composition analysis on the metal sheet based on the calibrated spectrum at each acquisition point.

[0006] The present invention has the following beneficial effects: Based on the above technical solution, this application can acquire detection data for each collection point of the metal sheet during the current detection and multiple historical detections. For each collection point, based on the degree of difference between the real-time operating condition data of the collection point during the current detection and multiple historical detections, at least one similar historical detection is determined for the collection point from multiple historical detections. Then, for each collection point, the reference spectrum of the current detection collection point is corrected based on the reference spectrum of the collection point during at least one similar historical detection, resulting in the corrected spectrum of the collection point. Finally, the chemical composition analysis of the metal sheet is performed based on the corrected spectrum of each collection point. The above technical solution fully considers the influence of dynamic changes in operating conditions during the production of metal sheets on the spectral signal. By screening historical detection data similar to the current detection conditions through operating condition similarity matching, and using the spectral data of similar historical detections to dynamically correct the current detection spectrum, the systematic spectral drift caused by changes in operating conditions such as temperature gradient, sheet warping, and surface oxidation is effectively eliminated, providing a more accurate and reliable spectral basis for chemical composition analysis, thereby significantly improving the accuracy and stability of online detection of chemical composition of metal sheets. Attached Figure Description

[0007] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0008] Figure 1 This is a schematic flowchart of a method for analyzing the chemical composition of thin metal plates based on spectral analysis, provided in one embodiment of the present invention. Figure 2 This is a system architecture diagram of a metal thin plate chemical composition analysis system based on spectral analysis, provided as an embodiment of the present invention. Detailed Implementation

[0009] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of the spectral analysis-based chemical composition analysis method and system for thin metal plates proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0010] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0011] In all division and logarithmic operations covered in this application, a smoothing mechanism is employed to prevent computer program crashes or invalid values ​​from being generated due to a zero denominator or a zero input. Specifically, a positive correction factor is superimposed on the denominator term of the division operation or the argument term of the logarithmic function. For example, the value is This ensures the robustness and feasibility of the algorithm under extreme conditions.

[0012] The normalization function mentioned in this application Unless otherwise specified, all values ​​are normalized using maximum and minimum values. The maximum and minimum values ​​are preset empirical extreme values ​​derived from a large amount of historical experimental data. If the calculated result exceeds the [0,1] interval, it is restricted to the [0,1] range by a truncation function (i.e., if the result is less than 0, it is taken as 0, and if it is greater than 1, it is taken as 1) to eliminate the influence of outliers on the evaluation index.

[0013] The specific scheme of the chemical composition analysis method and system for thin metal plates based on spectral analysis provided by the present invention will be described in detail below with reference to the accompanying drawings.

[0014] Please see Figure 1 The diagram illustrates a flowchart of a method for analyzing the chemical composition of thin metal plates based on spectral analysis, according to an embodiment of the present invention. The method includes the following steps: Step 101: Obtain the detection data for each collection point of the metal sheet during the current detection and multiple historical detections.

[0015] The detection data includes real-time operating data and reference spectra of the corresponding collection points.

[0016] In the above scheme, each collection point of the metal sheet refers to a discrete point set when collecting spectral and working condition data on the surface of the metal sheet during the detection process. The distribution of the collection points can be reasonably set according to the size of the metal sheet and the detection accuracy requirements. For example, the collection points can be evenly distributed along the width of the sheet and collected synchronously and continuously along the length at the production line speed to achieve full coverage of the metal sheet.

[0017] Real-time operating condition data refers to parameters related to the production environment and sheet condition at the collection point during testing. It can reflect the dynamic environmental characteristics during testing and provide a basis for subsequent matching of similar historical tests. For example, real-time operating condition data includes, but is not limited to, real-time temperature, thickness distribution, running tension, surface oxide film thickness, running speed, and other data of metal thin sheets. These data can comprehensively characterize the operating condition during testing.

[0018] A reference spectrum refers to a spectral signal acquired by a spectral detection device that reflects the chemical composition characteristics of a thin metal sheet. It includes reflectance or emission spectral curves and contains characteristic spectral information of each element in the thin metal sheet. It is the core basic data for chemical composition analysis.

[0019] In some embodiments, the current inspection data can be collected in real time by inspection equipment deployed on the production line, and the historical inspection data can be retrieved from the Manufacturing Execution System (MES). When retrieving the data, it is necessary to ensure that the historical data is consistent with the type of metal sheet and the production process range of the current inspection to ensure the effectiveness of matching similar historical inspections.

[0020] For example, this application can employ a high-speed linear array hyperspectral camera to continuously scan along the width of the metal sheet, acquiring the reflection / emission spectral curves of each pixel at different wavelengths. The acquisition frequency is synchronized with the production line speed (for example, once per second), achieving comprehensive coverage of the continuous strip. Laser-induced breakdown spectroscopy (LIBS) probes are deployed at key points on the production line to perform micro-area ablation on the sheet, acquiring plasma emission spectra containing elemental characteristic peaks as calibration and verification benchmarks for hyperspectral data. The acquisition frequency is once per second. Real-time temperature field, thickness distribution, tension, and operating speed data of the sheet are acquired based on production line sensors (such as temperature sensors, thickness sensors, tension sensors, and speed sensors). Based on the strip's operating speed and sensor distance, the operating data is interpolated or time-delayed to ensure data time sequence consistency. A communication connection is established with the MES system to acquire historical production process parameter data and laboratory chemical composition analysis reports for corresponding batches of products.

[0021] Step 102: For each collection point, based on the degree of difference between the real-time operating data of the collection point during the current detection and multiple historical detections, determine at least one similar historical detection for the collection point from multiple historical detections.

[0022] It should be noted that the spectral signal of thin metal sheets can drift due to changes in operating conditions. Direct comparison of spectral data under different operating conditions will introduce systematic errors. Therefore, it is necessary to select historical test data similar to the current operating conditions as a reference. This application can quantify the deviation between the current and historical operating conditions by analyzing the degree of difference between the real-time operating condition data collected at the current test and multiple historical tests. The smaller the difference, the closer the operating conditions are, and the higher the reference value of the corresponding historical reference spectrum for correcting the current spectrum.

[0023] In some embodiments, this application can calculate the degree of difference between the current real-time operating condition data and each historical real-time operating condition data through a preset difference evaluation algorithm, and then select historical detections that meet the preset similarity conditions as similar historical detections based on the magnitude of the difference, thereby eliminating interference from historical data with excessively large operating condition differences, ensuring that the historical reference spectra used for subsequent correction are targeted and effective, and providing a guarantee for improving the accuracy of spectral correction.

[0024] Step 103: For each acquisition point, based on the reference spectrum of at least one similar historical acquisition point, correct the reference spectrum of the current acquisition point to obtain the corrected spectrum of the acquisition point.

[0025] The corrected spectrum refers to the spectral signal after eliminating the systematic spectral drift caused by changes in operating conditions, which is closer to the true spectral characteristics of the thin metal plate under standard operating conditions. Since the operating conditions of similar historical tests are similar to the current operating conditions, the pattern of interference from operating conditions on the reference spectrum is consistent with the current spectrum. Therefore, this application can analyze the spectral differences between the two to remove the drift components caused by changes in operating conditions from the current spectrum.

[0026] In this way, the present application can use the spectral characteristics of similar historical data to reverse the systematic error of the current spectrum, avoid the distortion of spectral signals caused by dynamic changes in operating conditions, and thus provide an accurate spectral basis for subsequent chemical composition analysis.

[0027] Step 104: Perform chemical composition analysis on the metal sheet based on the calibrated spectrum at each collection point.

[0028] Since the calibrated spectrum has eliminated the interference of operating conditions, it can truly reflect the chemical composition characteristics of the metal sheet. Therefore, component analysis based on the calibrated spectrum can significantly improve the accuracy of the results.

[0029] In one possible implementation, this application can extract multiple spectral features from the calibration spectrum of each acquisition point to form a spectral feature vector of the acquisition point.

[0030] Among them, spectral features include at least one of elemental fingerprint features, spectral morphology features, relation ratio features, and spatial neighborhood features.

[0031] Elemental fingerprint characteristics refer to parameters such as spectral intensity and peak area of ​​each element in a thin metal sheet at characteristic wavelengths. These are the most direct basis for quantitative analysis of chemical composition, directly reflecting the elemental content. Spectral morphology characteristics refer to parameters such as the overall shape, peak shape, and baseline trend of the spectral curve. These are sensitive to changes in the phase composition and grain size of the thin metal sheet's microstructure, indirectly reflecting the distribution characteristics of chemical components. Relationship ratio characteristics refer to parameters such as the intensity ratio and peak area ratio at different characteristic wavelengths. These can eliminate some systematic errors (such as fluctuations in light source intensity) and improve the stability of the characteristics. Spatial neighborhood characteristics refer to parameters such as the correlation and gradient changes of the spectral characteristics between the current acquisition point and surrounding acquisition points, reflecting the spatial distribution trend of chemical components.

[0032] For example, this application can extract the above features from the calibration spectrum using feature extraction algorithms (such as wavelet transform, principal component analysis, etc.), and arrange the extracted features in a preset order to form the spectral feature vector of the acquisition point (e.g., with a dimension of M×1, where M is the total number of extracted features).

[0033] Then, the spectral feature vector of each collection point is input into the chemical composition prediction model to obtain the predicted chemical composition value of each collection point.

[0034] Among them, the chemical composition prediction model is a regression model trained on historical data. For example, a partial least squares regression model can be used. This model can effectively handle the multicollinearity problem between spectral features and chemical composition, and improve prediction accuracy.

[0035] The model training process is as follows: using the spectral feature vectors of each collection point in the historical detection as independent variables (forming a feature matrix), and the corresponding laboratory-verified chemical composition values ​​as dependent variables (forming a composition matrix), the least squares regression model is obtained by decomposing the two matrices. After training, the model is validated and optimized to ensure that the prediction error of the model is within the preset range (e.g., relative error less than 3%).

[0036] Thus, after inputting the spectral feature vector of the current collection point into the trained chemical composition prediction model, the chemical composition prediction model outputs the predicted values ​​of the chemical composition of each element at the collection point (such as the mass fraction of elements such as iron, carbon, and manganese).

[0037] In some embodiments, this application may also perform data fusion using a spatial interpolation algorithm based on the predicted chemical composition of each collection point and the positional distribution of each collection point on the metal sheet to obtain a two-dimensional distribution map of the chemical composition of the metal sheet.

[0038] Among them, the two-dimensional distribution map of chemical composition is used to characterize the spatial distribution of the content of each element on the surface of the thin metal plate.

[0039] Spatial interpolation algorithms are used to generate continuous chemical composition distributions on the surface of thin metal plates based on predicted values ​​from discrete acquisition points. For example, the spatial interpolation algorithm can be co-kriging interpolation, which can fuse predicted data from hyperspectral pixels and high-precision verification data from LIBS points, while also considering spatial neighborhood features to improve the accuracy of the interpolation results.

[0040] For example, firstly, the coordinate position of each collection point on the metal sheet is recorded (e.g., the x-axis is the width direction and the y-axis is the length direction). The coordinates of the collection points are associated with the corresponding predicted chemical composition values. Then, the coordinate region of the entire metal sheet is interpolated using the co-kriging interpolation method to obtain the chemical composition content of each coordinate point, and finally, a two-dimensional distribution map of chemical composition is generated.

[0041] This two-dimensional distribution map can intuitively show the spatial distribution differences of the content of various elements on the surface of a thin metal sheet, calculate the standard deviation of the content of each element and the proportion of segregation area, identify key weak elements (i.e. elements whose content exceeds the preset standard range), and provide intuitive basis for production process optimization and product quality grading.

[0042] Based on the above technical solution, this application can acquire detection data for each collection point of the metal sheet during the current detection and multiple historical detections. For each collection point, based on the degree of difference between the real-time operating condition data of the collection point during the current detection and multiple historical detections, at least one similar historical detection is determined for the collection point from multiple historical detections. Then, for each collection point, the reference spectrum of the current detection collection point is corrected based on the reference spectrum of the collection point during at least one similar historical detection, resulting in the corrected spectrum of the collection point. Finally, the chemical composition analysis of the metal sheet is performed based on the corrected spectrum of each collection point. The above technical solution fully considers the influence of dynamic changes in operating conditions during the production of metal sheets on the spectral signal. By screening historical detection data similar to the current detection conditions through operating condition similarity matching, and using the spectral data of similar historical detections to dynamically correct the current detection spectrum, the systematic spectral drift caused by changes in operating conditions such as temperature gradient, sheet warping, and surface oxidation is effectively eliminated, providing a more accurate and reliable spectral basis for chemical composition analysis, thereby significantly improving the accuracy and stability of online detection of chemical composition of metal sheets.

[0043] As a possible embodiment of this application, step 102 above can be implemented through the following steps: Step 201: For each collection point, calculate the difference in operating conditions between the real-time operating condition data of the collection point at the current detection time and the real-time operating condition data of the collection point at each historical detection time.

[0044] The operating condition difference degree characterizes the deviation between the current operating condition and the historical operating condition at the data collection point. A smaller operating condition difference degree indicates that the current operating condition is closer to the historical operating condition. Since real-time operating condition data contains parameters in multiple dimensions (such as temperature, thickness, tension, etc.), the operating condition parameters of each dimension can be standardized first (eliminating dimensional differences), and then the current operating condition data vector and the historical operating condition data vector can be constructed. The operating condition difference degree is then obtained through vector difference calculation methods.

[0045] For example, Euclidean distance can be used to calculate vector differences, and the currently detected real-time operating condition data vector is denoted as... (i.e., the first) The current detection parameters of each dimension at each collection point), the first The first collection point The real-time operating condition data vector of the previous historical test is denoted as: The difference in operating conditions is... and The Euclidean distance.

[0046] It should be noted that the calculation method for working condition difference is not limited to Euclidean distance. Other vector difference assessment methods such as Manhattan distance and cosine similarity can also be used. The specific method can be selected according to the distribution characteristics of the working condition parameters and the detection accuracy requirements.

[0047] Step 202: Select at least one similar historical detection from multiple historical detections based on the degree of difference in working conditions.

[0048] After obtaining the difference in operating conditions between the current detection and each historical detection, similar historical detections can be filtered using preset filtering rules. For example, a difference threshold can be preset (e.g., 0.2, the specific value can be calibrated according to the actual production scenario), and historical detections with an operating condition difference less than this threshold can be regarded as similar historical detections. Alternatively, filtering can be performed based on the relative magnitude of the difference, selecting one or more historical detections that are closest to the current operating condition, ensuring that the selected historical detections can truly reflect the similarity to the current operating condition.

[0049] In one possible implementation, multiple historical detections are sorted in ascending order according to the degree of difference in working conditions to obtain a historical detection sequence. Then, a preset number of historical detections that rank first in the historical detection sequence are taken as at least one similar historical detection.

[0050] By sorting in ascending order, historical tests with the smallest operating condition differences (i.e., most similar to the current operating condition) are placed at the beginning of the sequence, while those with larger differences are placed at the end. This visually presents the priority of the similarity between each historical test and the current operating condition. The preset number can be determined based on a combination of testing accuracy and computational efficiency, ranging from 3 to 10. For example, a value of 5 might be insufficient, affecting calibration accuracy, while a value that is too large would increase computation and reduce testing efficiency. The preset number can be adjusted for calibration based on factors such as the type of metal sheet and the range of operating condition fluctuations in actual production.

[0051] Based on the above technical solution, this application calculates the difference in operating conditions between the real-time operating data of the current detection point and the real-time operating data of the historical detection points for each collection point. Then, based on the difference in operating conditions, at least one similar historical detection is selected from multiple historical detections. This technical solution quantifies the difference in operating conditions based on operating parameters, objectively assessing the similarity between the current and historical detections. By selecting the historical detection with the smallest difference in operating conditions, it ensures that the historical data used for spectral correction is highly similar to the current detection in terms of operating conditions. This provides a reliable reference benchmark for subsequent spectral correction, avoids spectral correction deviations caused by excessive differences in operating conditions, and improves the targeting and effectiveness of spectral correction.

[0052] As a possible embodiment of this application, step 103 above can be implemented through the following steps: Step 301: For each acquisition point, determine the spectral correction weight corresponding to each similar historical detection based on the difference between the reference spectrum of the acquisition point at the current detection time and the reference spectrum of the acquisition point at each similar historical detection time.

[0053] The spectral correction weight is used to characterize the correction contribution of the reference spectrum of the corresponding similar historical detection to the current detection's reference spectrum. It can reflect the correction value of similar historical reference spectra. The larger the spectral correction weight, the greater the correction contribution of the historical reference spectrum to the current spectrum. Since the spectral differences between different similar historical detections and the current detection are different, their correction values ​​also differ. Therefore, it is necessary to determine the weight through spectral difference analysis to make the correction process more targeted.

[0054] For example, spectral differences can be evaluated from multiple dimensions such as characteristic wavelength matching degree and full spectrum consistency. The better the evaluation result (i.e., the smaller the spectral difference), the greater the corresponding spectral correction weight, ensuring that the correction process mainly relies on historical data that best matches the current spectrum.

[0055] Step 302: Perform weighted correction on the reference spectrum of the current detection point based on the spectral correction weight to obtain the corrected spectrum of the acquisition point.

[0056] In one possible implementation, this application can perform weighted processing on the reference spectrum of the acquisition point during each similar historical detection according to the spectral correction weight corresponding to each similar historical detection, to obtain the spectral correction amount corresponding to each similar historical detection. Then, the reference spectrum of the acquisition point during the current detection is corrected according to the spectral correction amount corresponding to each similar historical detection to obtain the corrected spectrum.

[0057] For example, the corrected spectrum satisfies the following formula: in, For the first Corrected spectra at each acquisition point For the current detection time The reference spectrum of each collection point This represents the total number of similar historical detections. For the first The first collection point The spectral correction weights corresponding to the next similarity history detection. For the first The first time similar historical detection The reference spectrum at each sampling point is represented by a reference drift vector relative to the standard conditions (i.e., the drift component in the historical spectrum caused by operating conditions). For example, this application can establish a standard reference spectrum by acquiring the spectrum of a standard sample under standard conditions. For each historical detection, the reference drift vector is determined based on the difference between the corresponding reference spectrum and the standard reference spectrum.

[0058] Characterizing the first The spectral correction amount corresponding to the next similar historical detection. The comprehensive correction value, derived from the spectral correction values ​​corresponding to all similar historical detections, can comprehensively and accurately reflect the drift pattern under current operating conditions. This comprehensive correction value is used to correct the current reference spectrum, ensuring that the corrected spectrum... It is closer to the true spectrum of a thin metal sheet under standard operating conditions.

[0059] Based on the above technical solution, this application determines the spectral correction weight for each similar historical detection based on the difference between the reference spectrum of the current detection point and the reference spectrum of the detection point in each similar historical detection for each collection point. Then, the reference spectrum of the current detection point is weighted and corrected based on the spectral correction weight to obtain the corrected spectrum of the collection point. The above technical solution enables the spectral correction process to be dynamically adjusted according to the correction contribution of different historical data, avoiding the correction deviation caused by treating all similar historical data equally. Through weighted correction, the advantages of multiple similar historical data can be integrated to more accurately remove systematic drift in the current spectrum, further improving the accuracy of the corrected spectrum and providing a more reliable basis for subsequent chemical composition analysis.

[0060] As a possible embodiment of this application, step 301 above can be implemented through the following steps: Step 401: For each acquisition point, determine the matching degree of the reference spectrum between the current acquisition point and the reference spectrum of each similar historical acquisition point based on the difference between the reference spectrum of the acquisition point at the current detection time and the reference spectrum of the acquisition point at each similar historical acquisition time.

[0061] Among them, the reference spectrum matching degree is used to characterize the similarity between the reference spectrum of the current detection point and the reference spectrum of similar historical detection points. The smaller the difference between the reference spectrum of the current detection point and the reference spectrum of the similar historical detection point, the higher the reference spectrum matching degree, indicating that the spectral characteristics of the two are closer, the higher the correction reference value of the corresponding historical spectrum for the current spectrum, and the greater its spectral correction weight.

[0062] In one possible implementation, this application can calculate, for each acquisition point, the initial difference in characteristic wavelength between the reference spectrum of the acquisition point at the current detection time and the reference spectrum of the acquisition point at each similar historical detection time.

[0063] Among them, characteristic wavelengths are used to characterize the reaction wavelengths of the chemical composition of the metal sheet in the spectrum. For example, the emission or absorption wavelengths corresponding to each element in the metal sheet can be determined through prior correlation analysis. The number of characteristic wavelengths can be determined according to the type of element to be detected.

[0064] For example, the initial difference values ​​satisfy the following formula: in, For the current detection time The reference spectrum of the first collection point and the first The first time similar historical detection The initial difference values ​​of the reference spectrum at the characteristic wavelength at each acquisition point The number of characteristic wavelengths, For the current detection time The reference spectrum of the acquisition point is the th Intensity values ​​at each characteristic wavelength For the first The first time similar historical detection The reference spectrum of the acquisition point is the th Intensity values ​​at each characteristic wavelength.

[0065] Initial difference value The smaller the value, the higher the value at the current detection time. The reference spectrum of the first collection point and the first The first time similar historical detection The smaller the intensity difference of the reference spectrum at the characteristic wavelength among the collection points, the higher the spectral similarity.

[0066] Then, for each acquisition point, a stability index is calculated between the reference spectrum of the acquisition point at the current detection time and the reference spectrum of the acquisition point at each similar historical detection time.

[0067] Among them, the stability index is used to characterize the stability of the overall difference between the reference spectrum of the current detection point and the reference spectrum of the corresponding similar historical detection points. It can assess the overall consistency between the current spectrum and the historical spectrum across the entire spectrum range, including the overall waveform shift and distortion.

[0068] For example, the stability index satisfies the following formula: in, For the current detection time The reference spectrum of the first collection point and the first The first time similar historical detection Stability index among the reference spectra of each acquisition point For the current detection time The reference spectrum of the first collection point and the first The first time similar historical detection The intensity difference vector of the reference spectrum at each wavelength across the full spectrum at each sampling point. This is the average of the absolute values ​​of all elements in the intensity difference vector, representing the overall shift of the spectral waveform. A smaller value indicates a smaller overall shift. The standard deviation of each element in the intensity difference vector represents the degree of distortion of the spectral waveform; the smaller the value, the more uniform the waveform. These are weighting coefficients used to adjust the influence weights of overall offset and distortion, satisfying... ,For example, It is 0.7. The value is 0.3, which can be calibrated according to the actual testing scenario.

[0069] The larger the value, the smaller the overall difference between the current spectrum and the historical spectrum across the entire spectrum, the higher the stability, and the better the spectral similarity.

[0070] Thus, this application can determine the baseline spectral matching degree between the current detection and each similar historical detection based on the initial difference value and stability index.

[0071] For example, the reference spectral matching degree satisfies the following formula: in, For the first The current detection corresponding to the collection point and the first... The baseline spectral matching degree of the second similarity historical detection For the current detection time The reference spectrum of the first collection point and the first The first time similar historical detection Stability index among the reference spectra of each acquisition point For the current detection time The reference spectrum of the first collection point and the first The first time similar historical detection The initial difference values ​​of the reference spectrum at the characteristic wavelength at each acquisition point It is a safety parameter used to correct fractions where the denominator is 0, and its dimensions are the same as... The same applies; the specific value can be determined based on... The value of the value determines the outcome, such as . This is a normalization function (e.g., maximum / minimum normalization) used to map the calculation results to the range of 0 to 1.

[0072] Step 402: Determine the spectral correction weights corresponding to each similar historical detection based on the baseline spectral matching degree.

[0073] For example, the spectral correction weights satisfy the following formula: in, For the first The first collection point The spectral correction weights corresponding to the next similarity history detection. For the first The current detection corresponding to the collection point and the first... The baseline spectral matching degree of the second similarity historical detection This represents the total number of similar historical detections. For the first The current detection corresponding to the collection point and the first... The baseline spectral matching degree of similar historical detection. The above formula, which represents the sum of the baseline spectral matching degree of all similar historical detections, converts the baseline spectral matching degree into spectral correction weights. It can directly associate the matching degree with the correction weights. The higher the matching degree of historical detections, the greater their correction weights, ensuring that the correction process preferentially relies on historical data with higher spectral similarity.

[0074] Based on the above technical solution, this application can determine the reference spectral matching degree between the current detection and each similar historical detection for each acquisition point, according to the difference between the reference spectrum of the acquisition point at the time of the current detection and the reference spectrum of the acquisition point at each similar historical detection. Then, the spectral correction weight corresponding to each similar historical detection is determined based on the reference spectral matching degree. This adaptively utilizes information from multiple similar historical detections, avoiding the random errors that may exist in a single historical detection, ensuring the dominant role of high-matching historical detections, and improving the robustness and accuracy of spectral correction.

[0075] Please see Figure 2 The diagram illustrates a system architecture of a metal thin-plate chemical composition analysis system based on spectral analysis, according to an embodiment of the present invention. The metal thin-plate chemical composition analysis system 20 based on spectral analysis includes: The data acquisition module 21 is used to acquire the detection data of each collection point of the metal sheet during the current detection and multiple historical detections; the detection data includes the real-time operating data and reference spectrum of the corresponding collection point; The historical matching module 22 is used to determine at least one similar historical detection for each collection point based on the degree of matching between the real-time operating data of the collection point during the current detection and multiple historical detections. Spectral correction module 23 is used to correct the reference spectrum of the current detection point for each acquisition point based on the reference spectrum of at least one similar historical acquisition point, so as to obtain the corrected spectrum of the acquisition point. The component analysis module 24 is used to perform chemical composition analysis on the metal sheet based on the calibrated spectrum of each acquisition point.

[0076] It should be noted that the various embodiments of this application can be referenced or learned from each other. For example, the same or similar steps, method embodiments, system embodiments and device embodiments can be referenced from each other without limitation.

[0077] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0078] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A method for chemical composition analysis of thin metal plates based on spectral analysis, characterized in that, include: The detection data for each collection point of the metal sheet is acquired during the current detection and multiple historical detections; the detection data includes real-time operating data and reference spectrum of the corresponding collection point. For each collection point, based on the degree of difference between the real-time operating data of the collection point during the current detection and multiple historical detections, at least one similar historical detection is determined for the collection point from the multiple historical detections; For each acquisition point, the reference spectrum of the acquisition point at the current detection time is corrected based on the reference spectrum of the acquisition point at the at least one similar historical detection time to obtain the corrected spectrum of the acquisition point; The chemical composition of the metal sheet is analyzed based on the calibrated spectrum at each collection point.

2. The method for chemical composition analysis of thin metal plates based on spectral analysis according to claim 1, characterized in that, For each data collection point, based on the degree of difference between the real-time operating data of the data collection point during the current detection and multiple historical detections, at least one similar historical detection is determined for the data collection point from the multiple historical detections, including: For each data collection point, the operating condition difference degree between the real-time operating condition data of the data collection point at the current detection time and the real-time operating condition data of the data collection point at each historical detection time is calculated; the operating condition difference degree is used to characterize the degree of deviation between the operating condition state of the data collection point at the current detection time and the operating condition state at the historical detection time. Based on the difference in operating conditions, at least one similar historical detection is selected from the multiple historical detections.

3. The method for chemical composition analysis of thin metal plates based on spectral analysis according to claim 2, characterized in that, Based on the difference in operating conditions, at least one similar historical detection is selected from the multiple historical detections, including: The historical detections are sorted in ascending order according to the difference in operating conditions to obtain a historical detection sequence. The preset number of historical detections that rank highest in the historical detection sequence are taken as the at least one similar historical detection.

4. The method for chemical composition analysis of thin metal plates based on spectral analysis according to claim 1, characterized in that, For each acquisition point, based on the reference spectrum of the acquisition point during at least one similar historical detection, the reference spectrum of the acquisition point during the current detection is corrected to obtain the corrected spectrum of the acquisition point, including: For each acquisition point, the spectral correction weight corresponding to each similar historical detection is determined based on the difference between the reference spectrum of the acquisition point at the current detection and the reference spectrum of the acquisition point at each similar historical detection. The spectral correction weight is used to characterize the degree of correction contribution of the reference spectrum of the corresponding similar historical detection to the reference spectrum of the current detection. The reference spectrum of the acquisition point at the current detection time is weighted and corrected based on the spectral correction weights to obtain the corrected spectrum of the acquisition point.

5. The method for chemical composition analysis of thin metal plates based on spectral analysis according to claim 4, characterized in that, For each acquisition point, based on the difference between the reference spectrum of the acquisition point at the current detection time and the reference spectrum of the acquisition point at each similar historical detection time, the spectral correction weight corresponding to each similar historical detection is determined, including: For each acquisition point, the matching degree of the reference spectrum between the current acquisition point and each similar historical acquisition point is determined based on the difference between the reference spectrum of the acquisition point at the time of the current acquisition and the reference spectrum of the acquisition point at each similar historical acquisition point. The spectral correction weights corresponding to each similar historical detection are determined based on the baseline spectral matching degree.

6. The method for chemical composition analysis of thin metal plates based on spectral analysis according to claim 5, characterized in that, For each acquisition point, the matching degree of the reference spectrum between the current acquisition point and each similar historical acquisition point is determined based on the difference between the reference spectrum of the acquisition point at the time of the current acquisition and the reference spectrum of the acquisition point at each similar historical acquisition point, including: For each acquisition point, the initial difference value at a characteristic wavelength between the reference spectrum of the acquisition point at the current detection time and the reference spectrum of the acquisition point at each similar historical detection time is calculated. The characteristic wavelength is used to characterize the reaction wavelength of the chemical composition of the metal sheet in the spectrum. For each acquisition point, a stability index is calculated between the reference spectrum of the acquisition point at the current detection time and the reference spectrum of the acquisition point at each similar historical detection time; the stability index is used to characterize the stability of the overall difference between the reference spectrum of the acquisition point at the current detection time and the reference spectrum of the acquisition point at the corresponding similar historical detection time. The baseline spectral matching degree between the current detection and each similar historical detection is determined based on the initial difference value and the stability index.

7. The method for chemical composition analysis of thin metal plates based on spectral analysis according to claim 4, characterized in that, The reference spectrum of the acquisition point at the current detection time is weighted and corrected based on the spectral correction weights to obtain the corrected spectrum of the acquisition point, including: Based on the spectral correction weight corresponding to each similar historical detection, the reference spectrum of the collection point during each similar historical detection is weighted to obtain the spectral correction amount corresponding to each similar historical detection. The reference spectrum of the acquisition point at the current detection time is corrected based on the spectral correction amount corresponding to each similar historical detection to obtain the corrected spectrum.

8. The method for chemical composition analysis of thin metal plates based on spectral analysis according to claim 1, characterized in that, The chemical composition of the thin metal plate was analyzed based on the calibrated spectrum at each acquisition point, including: For each acquisition point, multiple spectral features are extracted from the calibration spectrum of the acquisition point to form the spectral feature vector of the acquisition point; the spectral features include at least one of elemental fingerprint features, spectral morphology features, relation ratio features, and spatial neighborhood features. The spectral feature vector of each collection point is input into the chemical composition prediction model to obtain the predicted chemical composition value of each collection point.

9. The method for chemical composition analysis of thin metal plates based on spectral analysis according to claim 8, characterized in that, The method further includes: Based on the predicted chemical composition values ​​of each collection point and the location distribution of each collection point on the metal sheet, a two-dimensional distribution map of the chemical composition of the metal sheet is obtained by data fusion through a spatial interpolation algorithm; the two-dimensional distribution map of the chemical composition is used to characterize the spatial distribution of the content of each element on the surface of the metal sheet.

10. A system for analyzing the chemical composition of thin metal plates based on spectral analysis, characterized in that, include: The data acquisition module is used to acquire the detection data of each collection point of the metal sheet during the current detection and multiple historical detections; the detection data includes the real-time operating condition data and reference spectrum of the corresponding collection point; The historical matching module is used to determine at least one similar historical detection for each collection point based on the degree of matching between the real-time operating data of the collection point during the current detection and multiple historical detections. A spectral correction module is used to correct the reference spectrum of the acquisition point at the current detection time based on the reference spectrum of the acquisition point at the at least one similar historical detection time for each acquisition point, so as to obtain the corrected spectrum of the acquisition point. The component analysis module is used to perform chemical composition analysis on the metal sheet based on the calibrated spectrum at each acquisition point.

Images