Temperature time-varying molten steel LIBS spectrum modeling method in RH refining process

Abstract

The invention provides a temperature time-varying molten steel LIBS spectrum modeling method in the RH refining process, and belongs to the field of steel smelting. The method comprises the following steps: firstly, acquiring LIBS spectral data of molten steel, selecting n element modeling analysis lines from the spectral data, and for the n element modeling analysis lines, calculating a spectral intensity mean value and a variance of N times of excitation of a single analysis line; selecting an analysis line corresponding to the spectral intensity value of the minimum variance excited for N times in the n analysis lines as a starting point analysis line, drawing a slope curve of the normal temperature T0 and the corresponding high temperature T with the other analysis lines, and then calculating the spectral intensity values of the n analysis lines at the normal temperature and the corresponding high temperature; deriving an exponential relationship between the slope and the spectral intensity along with the temperature change, further deriving a sensitivity coefficient of the slope and the spectral intensity to the temperature, and weighting and correcting element concentration predicted values at different temperatures through a regression coefficient according to a sensitivity coefficient ai of the slope to the temperature and a sensitivity coefficient bi of the spectral intensity to the temperature; the smelting detection efficiency is improved.

Description

Technical field:

[0001] The present invention belongs to the field of steel smelting, and in particular relates to a LIBS spectral modeling method for time-varying temperature of molten steel in a RH refining process. Background technology:

[0002] Steelmaking is a general term for the metallurgical processes of steel and iron. Steelmaking uses pig iron smelted in blast furnaces, sponge iron smelted by direct reduction ironmaking, and scrap steel as raw materials, and is made into steel by different methods. The basic production process is to smelt iron ore into pig iron in a blast furnace, and then use pig iron as raw material to smelt steel by different methods, and then cast it into ingots or continuous casting billets; the refining process is an important step in the steelmaking process. Refining plays the role of deoxidation, decarbonization, desulfurization, removal of impurities, modification of inclusions, and adjustment of the composition of molten steel. It is an important link to improve the quality of steel products, smelt high-end steel grades, optimize production process flow, improve production efficiency and ensure stable and smooth continuous casting; the endpoint control of the refining process is a key control step related to the quality of steel products, and its ultimate goal is to achieve fast and accurate detection of molten steel composition.

[0003] In the existing technology, the composition detection of the refining process adopts offline and intermittent detection, and real-time synchronous detection is not achieved. The composition adopts a prediction model, resulting in a lack of real-time and accuracy in process control, and there are problems such as low control precision, low endpoint hit rate, low degree of intelligence, and great control difficulty. Summary of the invention:

[0004] To address the above-mentioned issues, the present invention provides a spectral modeling method for laser-induced breakdown spectroscopy (LIBS) of molten steel with time-varying temperature during RH refining. This method utilizes a pulsed laser, an ICCD spectrometer, a digital time-delay pulse generator, and a long-distance optical path system to achieve online, rapid, and accurate detection of the elemental composition of molten steel with time-varying temperature during RH refining through optical fiber transmission, photoelectric conversion, and online analysis of spectral data.

[0005] In order to achieve the above objectives, the technical solutions adopted in the embodiments of the present invention are as follows:

[0006] A LIBS spectral modeling method for molten steel with time-varying temperature during RH refining comprises the following steps:

[0007] Step S1, obtaining LIBS spectrum data of molten steel in an RH refining furnace; the spectrum data includes spectrum intensity matrix data of different spectral line wavelengths of several elements;

[0008] Step S2, selecting n modeling analysis line data of a certain element from the spectral intensity matrix data;

[0009] Step S3, for the n modeling analysis lines of a certain element, respectively calculate the mean and variance of the spectral intensity of each analysis line N times of excitation;

[0010] Step S4: compare n analysis lines and select the spectral intensity value with the minimum variance of N excitations The corresponding analysis line is used as the starting analysis line λ′;

[0011] Step S5: Taking the starting analysis line λ′ as the starting point, and drawing the slope curves |K1|, |K2|, ..., |K at room temperature T0 with the remaining (n-1) analysis lines. (n-1) |; corresponding to the high temperature T slope curve |K′1|,|K′2|,...,|K′ (n-1) |; The spectral intensity values ​​of n analysis lines at room temperature T0 are I1, I2, ..., I n , corresponding to the high temperature T spectrum intensity values ​​I′1, I′2, ..., I′ n ;

[0012] Step S6: According to the blackbody radiation spectrum distribution of Planck's law, when the temperature rises, the absolute value of the slope and the spectral intensity increases exponentially, and the exponential relationship between the slope and the temperature and the spectral intensity are obtained; and further, based on the spectral intensity values ​​corresponding to the room temperature T0 and the temperature T of the high-temperature molten steel liquid, the slope sensitivity coefficient a is derived. i And the sensitivity coefficient b of spectral intensity to temperature i ;

[0013] Step S7, based on the slope sensitivity coefficient a to temperature i And the sensitivity coefficient b of spectral intensity to temperature i , the predicted element concentration C(T) at different temperatures is corrected by weighting the regression coefficient.

[0014] As a preferred embodiment of the present invention, in step S2, the PLSR regression coefficient feature selection method is adopted, and the non-interference of the element spectral lines and the relatively high intensity persistent lines matched by the NIST database are considered, and finally n spectral lines are selected as analysis lines for modeling.

[0015] As a preferred embodiment of the present invention, when the band range of the ICCD spectrometer used to collect spectral data is 199-936nm, the corresponding original spectral data contains 26,937 wavelength spectral lines; when selecting analysis lines from the 26,937 wavelength spectral lines, first, 150 spectral lines with large absolute values ​​of regression coefficients are selected from the 26,937 wavelengths through the PLSR regression coefficient feature selection method, and then the spectral data of the 150 spectral lines are used to find the spectral line peak points using the second-order derivative extreme point method, and matched with the NIST database, and the threshold for matching the peak point with the element NIST database is set to 0.05. When the difference between the peak point and the value in the NIST database is less than 0.05, it is defaulted to be the analysis line of the element, otherwise it is not; at the same time, the element interference lines eliminate the interference of Fe, Cr and Ni metal element spectral lines, and finally, n spectral lines with high relative intensity values ​​are selected as analysis lines for modeling.

[0016] As a preferred embodiment of the present invention, in step S3, the average value of N excitations of a single analysis line is:

[0017]

[0018] The variance of N excitations of a single analysis line is:

[0019]

[0020] In formulas (1) and (2), I k is the spectral intensity value of a single analytical line; is the mean spectral intensity of N excitations of a single analytical line; is the spectral intensity variance of N excitations of a single analytical line; (k = 1, 2, ..., N).

[0021] As a preferred embodiment of the present invention, in step S4, the spectral intensity value with the minimum variance for:

[0022]

[0023] The corresponding analysis line is:

[0024]

[0025] In formula (4), λ′ is the minimum variance spectral intensity value Corresponding analysis line; (i=1,2,...,n).

[0026] As a preferred embodiment of the present invention, in step S6, the blackbody radiation spectrum distribution formula is as follows:

[0027]

[0028] In formula (5), I(v,T) is the radiation intensity at frequency v; h is Planck's constant; c is the speed of light; v is the frequency of radiation; T is the temperature;

[0029] The exponential relationship between the slope and temperature is:

[0030]

[0031] In formula (6), |K i (T0)| is the slope at room temperature T0; a i is a constant term, which is the sensitivity of the slope to temperature;

[0032] According to the corresponding slope values ​​at room temperature T0 and high temperature molten steel temperature T, the sensitivity coefficient value a of the slope to temperature is derived. i :

[0033]

[0034] The exponential relationship between spectral intensity and temperature is:

[0035]

[0036] In formula (8), I i+1 (T0) is the spectral intensity at room temperature T0; b i is a constant term, which represents the sensitivity of spectral intensity to temperature;

[0037] According to the corresponding spectral intensity values ​​at room temperature T0 and high temperature molten steel temperature T, the sensitivity coefficient value b of spectral intensity to temperature is derived. i :

[0038]

[0039] As a preferred embodiment of the present invention, the sensitivity coefficient value b of the spectrum intensity and slope to temperature is calculated. i and a i When the temperature is set to 25°C, the room temperature T0 is set to 25°C and the high temperature molten steel temperature T is set to 1500°C.

[0040] As a preferred embodiment of the present invention, in step S7, the element concentration prediction value C(T) is calculated as follows:

[0041]

[0042] Further optimizing formula (10) yields:

[0043]

[0044] In formulas (10) and (11), β i+1is the regression coefficient (or weight) of the i-th spectral line. The regression coefficient value can be solved by the PLSR algorithm through the multivariate regression equation of the normal temperature spectral intensity matrix independent variable I and the element concentration vector dependent variable C; (i=1,2,...,(n-1)); a i is the sensitivity coefficient of slope to temperature; b i is the sensitivity coefficient of spectral intensity to temperature; is the spectral intensity value with minimum variance for n analysis lines and N excitations; I′ i+1 (T) is the spectral intensity at different high temperatures T; |K′ i (T)| is the slope at different high temperatures T; I i+1 (T0) is the spectral intensity at room temperature T0; |K i (T0)| is the slope at room temperature T0.

[0045] The solution of the embodiment of the present invention has the following beneficial effects:

[0046] The embodiment of the present invention provides a LIBS spectral modeling method for molten steel with time-varying temperature during RH refining. An exponential equation for spectral intensity and slope at different temperatures is established, and concentration values ​​at different temperatures are predicted using a weighted correction method using a regression coefficient. This method is suitable for LIBS composition detection under time-varying molten steel temperature. An industrial computer integrates a LIBS composition detection algorithm, software, and hardware system for long-distance optical paths to achieve real-time online LIBS composition prediction for molten steel, eliminating the need for offline pneumatic sampling, sample preparation, and manual detection steps. This saves labor costs, improves smelting detection efficiency, and achieves a high level of system integration and intelligence.

[0047] Of course, it is not necessary to achieve all of the advantages described above simultaneously in order to implement any product or method of the present invention. Description of the drawings:

[0048] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0049] Figure 1 Schematic diagram of the principle of the LIBS detection device for molten steel in the RH refining furnace according to an embodiment of the present invention;

[0050] Figure 2 Flowchart of the LIBS spectrum modeling method for temperature-varying molten steel during RH refining in an embodiment of the present invention;

[0051] Figure 3This is a slope curve diagram of LIBS element composition detection of molten steel at different temperatures in an embodiment of the present invention.

[0052] Description of reference numerals:

[0053] 1. Pulse laser; 2. Digital delayed pulse generator; 3. Industrial computer; 4. ICCD spectrometer; 5. Telescope system; 6. Vacuum induction furnace; 7. Molten steel; 8. Laser rangefinder; 9. Reflector. Specific implementation method:

[0054] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention; it is obvious that the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; the components of the embodiments of the present invention generally described and shown in the drawings here can be arranged and designed in various different configurations; it should be noted that the embodiments of the present invention and the features in the embodiments can also be combined with each other without conflict.

[0055] It should be noted that similar numbers and letters represent similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. In the description of the present invention, the terms "first", "second", "third", "fourth", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.

[0056] Based on the problem of detecting the composition of molten steel with time-varying temperature during the refining process of an RH refining furnace, an embodiment of the present invention provides a spectral modeling method for laser-induced breakdown spectroscopy (LIBS) of molten steel with time-varying temperature during the RH refining process, and adopts a long-distance optical path scheme to perform LIBS composition detection on the molten steel in the RH refining furnace; the detection method gives a spectral distribution formula of blackbody radiation according to Planck's law, and it can be seen that the spectral intensity increases exponentially with increasing temperature. First, the analysis line corresponding to the minimum variance intensity value of multiple excitations is used as the starting point, and other modeling analysis lines are connected to construct a slope curve. Then, exponential equations of the spectral intensity and slope at different temperatures are established respectively. Then, according to the spectral intensity values ​​and slope values ​​of the molten steel at room temperature of 25°C and 1500°C, the sensitivity coefficient values ​​b of the spectral intensity and slope to temperature are calculated respectively. i and a i Finally, the regression coefficient weighted correction method and the sensitivity coefficient value a i 、b i To predict the element concentration values ​​at different temperatures, and realize the LIBS spectrum modeling of the temperature-varying molten steel liquid during the RH refining process.

[0057] like Figures 1 to 2As shown, the RH refining process temperature time-varying molten steel LIBS spectrum modeling method provided by the embodiment of the application is based on Figure 1 As shown, the RH refining furnace molten steel LIBS detection device is modeled.

[0058] The device comprises a pulse laser 1, a digital delay pulse machine 2, an industrial computer 3, an ICCD spectrometer 4, a telescope system 5, a vacuum induction furnace 6, molten steel 7, a laser range finder 8 and a mirror 9. The pulse laser emits 1064nm single-wavelength pulse laser, the mirror reflects the horizontal laser into vertical downward laser, the laser is focused on the surface of the molten steel through the long-distance light path and the telescope system, and the focusing process is realized. The light signal is transmitted to the ICCD spectrometer through the telescope system in the light collection process, the photoelectric signal is converted, the spectral intensity and element concentration information are finally presented to the industrial computer, and the focusing and light collection processes of the whole LIBS light path are completed. The digital delay pulse machine is used for controlling the timing signals of the pulse laser and the ICCD spectrometer. The industrial computer is used for integrating the real-time LIBS composition information of the molten steel, and the RH refining process temperature time-varying molten steel end composition online prediction is realized through the software and algorithm integration.

[0059] Preferably, the energy of the pulse laser is 200mJ, the working voltage is 220V, the repetition frequency is 10Hz, the actual working frequency is 2.5Hz, the inherent delay time of the pulse laser is 110us, and the two signals of Q in and Clk in need to be received to trigger the laser. The digital delay pulse machine sets two channels, which are set to 100us and 110us respectively, and the voltage is set to 5V. The digital delay pulse machine is used for controlling the timing signals of the pulse laser and the ICCD spectrometer. The resolution of the ICCD spectrometer is lambda / 6000nm. The integration time is 1ms. The resolution of each channel is different. The wavelength range of the ICCD spectrometer is 199-936nm. The industrial computer is different from the ordinary computer and can work all year round. The molten steel is placed in the vacuum induction furnace, and the surface of the molten steel has no oxide layer. The single-wavelength of the laser range finder is 650nm, and the accuracy is 0.1mm. The laser range finder is used for real-time measurement of the focusing distance between the bottom end of the telescope system and the surface of the molten steel, which is 1550mm.

[0060] Based on the above device, the modeling method firstly takes the analysis line with the minimum variance spectral intensity value excited for multiple times as the starting point, connects other modeling analysis lines to construct a slope curve, then establishes an exponential equation of the spectral intensity and the slope at different temperatures, and finally predicts the concentration value at different temperatures through the regression coefficient weighting correction method to realize the LIBS spectrum modeling of the temperature time-varying molten steel.

[0061] As Figure 2 As shown, the specific steps include:

[0062] Step S1, obtaining LIBS spectrum data of molten steel in an RH refining furnace; the spectrum data includes spectrum intensity matrix data of different spectral line wavelengths of several elements;

[0063] Step S2, selecting n modeling analysis line data of a certain element from the spectral intensity matrix data;

[0064] In this step, n analysis lines are used for LIBS regression modeling. The partial least squares regression (PLSR) coefficient feature selection method is used, combined with the consideration of the absence of interference from elemental spectral lines and the relatively high intensity persistent lines matched by the NIST database. Finally, n spectral lines are selected as analysis lines for modeling.

[0065] In this embodiment, the ICCD spectrometer used captures a wavelength range of 199-936 nm, with 26,937 wavelength spectral lines. Selecting appropriate analytical line modeling from these 26,937 wavelength spectral lines is also a key step in LIBS quantitative analysis. 150 spectral lines with large absolute values ​​of regression coefficients are selected from the 26,937 wavelengths using the PLSR regression coefficient feature selection method. The spectral data of these 150 spectral lines are then used to find spectral line peaks using the second-order derivative extreme point method and matched with the NIST database. The threshold for matching the peak point with the element NIST database is set to 0.05. When the difference between the peak point and the value in the NIST database is less than 0.05, it is assumed to be the analytical line for the element, otherwise it is not. Secondly, the element interference lines should consider eliminating the interference of metal element spectral lines such as Fe, Cr, and Ni. Since Fe has the highest content, the NIST database of Fe element has more than 8,000 analytical lines and the most interference lines. After eliminating the interference lines, n spectral lines with high relative intensity values ​​are finally selected as analytical lines for modeling.

[0066] Step S3, for the n modeling analysis lines of a certain element, respectively calculate the mean and variance of the spectral intensity of each analysis line N times of excitation;

[0067] In this step, the mean value of N excitations for a single analysis line is:

[0068]

[0069] The variance of N excitations of a single analysis line is:

[0070]

[0071] In formulas (1) and (2), I k is the spectral intensity value of a single analytical line; is the average of spectral intensity of single analysis line N times of excitation; is the variance of spectral intensity of single analysis line N times of excitation; (k = 1, 2,..., N).

[0072] Step S4, comparing N analysis lines, selecting the spectral intensity value of minimum variance of N times of excitation The corresponding analysis line is taken as the starting analysis line λ';

[0073] In this step, the spectral intensity value of minimum variance is :

[0074]

[0075] The corresponding analysis line is:

[0076]

[0077] In formula (4), λ' is the spectral intensity value of minimum variance corresponding analysis line; (i = 1, 2,..., n).

[0078] Step S5, taking the starting analysis line λ' as the starting point, drawing the slope curve |K1|, |K2|,..., |K (n-1) | of normal temperature T0 with the remaining (n-1) analysis lines; drawing the slope curve |K'1|, |K'2|,..., |K' (n-1) | of high temperature T with the remaining (n-1) analysis lines; the spectral intensity values of n analysis lines of normal temperature T0 are I1, I2,..., In n , and the spectral intensity values of n analysis lines of high temperature T are I'1, I'2,..., I'n n ;

[0079] Step S6, according to the blackbody radiation spectral distribution of Planck's law, the absolute values of the slope and the spectral intensity exponentially increase when the temperature increases, the exponential relationship of the slope with the temperature and the exponential relationship of the spectral intensity with the temperature are obtained, and further, according to the spectral intensity values corresponding to normal temperature T0 and high temperature T of molten steel liquid, the temperature sensitivity coefficient a i of the slope to temperature and the temperature sensitivity coefficient b i of the spectral intensity to temperature are derived.

[0080] In this step, the formula of blackbody radiation spectral distribution is as follows:

[0081]

[0082] In formula (5), I(v, T) is the radiation intensity when the frequency is v; h is Planck's constant; c is the speed of light; v is the frequency of radiation; T is the temperature;

[0083] The exponential relationship between the slope and temperature is:

[0084]

[0085] In formula (6), |K i (T0)| is the slope at room temperature T0; a i is a constant term, which is the sensitivity of the slope to temperature;

[0086] According to the corresponding slope values ​​at room temperature T0 and high temperature molten steel temperature T, the sensitivity coefficient value a of the slope to temperature is derived. i :

[0087]

[0088] The exponential relationship between spectral intensity and temperature is:

[0089]

[0090] In formula (8), I i+1 (T0) is the spectral intensity at room temperature T0; b i is a constant term, which represents the sensitivity of spectral intensity to temperature;

[0091] According to the corresponding spectral intensity values ​​at room temperature T0 and high temperature molten steel temperature T, the sensitivity coefficient value b of spectral intensity to temperature is derived. i :

[0092]

[0093] In this step, preferably, the sensitivity coefficient value b of the spectral intensity and slope to temperature is calculated i and a i When the temperature is set to 25°C, the room temperature T0 is set to 25°C and the high temperature molten steel temperature T is set to 1500°C.

[0094] Step S7, based on the slope sensitivity coefficient a to temperature i And the sensitivity coefficient b of spectral intensity to temperature i , the predicted element concentration C(T) at different temperatures is corrected by weighting the regression coefficient.

[0095]

[0096] Further optimizing formula (10) yields:

[0097]

[0098] In formulas (10) and (11), β i+1is the regression coefficient (or weight) of the i-th spectral line. The regression coefficient value can be solved by the PLSR algorithm through the multivariate regression equation of the normal temperature spectral intensity matrix independent variable I and the element concentration vector dependent variable C; (i=1,2,...,(n-1)); a i is the sensitivity coefficient of slope to temperature; b i is the sensitivity coefficient of spectral intensity to temperature; is the spectral intensity value with minimum variance for n analysis lines and N excitations; I′ i+1 (T) is the spectral intensity at different high temperatures T; |K i ′(T)| is the slope at different high temperatures T; I i+1 (T0) is the spectral intensity at room temperature T0; |K i (T0)| is the slope at room temperature T0.

[0099] In this step, n analysis lines correspond to n spectral intensity values ​​and (n-1) slope values, so the spectral intensity value of the analysis line λ′ at the starting point of the slope is Extract separately and calculate the element concentration ratio value corresponding to the regression coefficient weight separately

[0100] The modeling method described in the embodiment of the present invention is applied to the detection device, and 12 samples at room temperature of 25°C with concentration labels and 8 samples at high temperature of 1500°C without concentration labels are selected for mixed modeling to perform component detection and analysis; first, the element modeling analysis line is selected; then the mean and variance of the spectral intensity of N excitations of 5 to 10 analysis lines are calculated, and the analysis line with the minimum variance intensity value of N excitations is used as the starting point, and the other modeling analysis lines are connected to construct a slope curve, and then the exponential equations of the spectral intensity and slope at room temperature of 25°C and high temperature of 1500°C are established respectively. At this time, the temperature sensitivity constant term a can be calculated based on the spectral intensity value and slope value at room temperature of 25°C and high temperature of 1500°C. i and b i Finally, the concentration values ​​at different temperatures are predicted by the regression coefficient weighted correction method to realize the component detection of variable temperature LIBS. Among them, the regression coefficient is obtained by the multivariate regression equation PLSR feature selection method.

[0101] Figure 3 The slope curve of LIBS element composition detection of molten steel at different temperatures in the embodiment of the present invention is shown in the figure. The horizontal axis is the wavelength and the vertical axis is the spectral intensity value. The 5 solid points in the figure represent the spectral intensity values ​​of the 5 analysis lines at room temperature and the spectral intensity value of the minimum variance of N excitations. The corresponding λ′ is the analysis line with the smallest variance among the five analysis lines. λ′ is the starting point, and the other solid points are connected to construct the slope curves K1, K2, ..., K4 at room temperature. The five hollow points in the figure represent the high-temperature spectral intensity values ​​of the five analysis lines. Since the temperature increases exponentially with the spectral intensity and slope, the high-temperature intensity value I' of the same analysis line λ′ is higher than the normal-temperature intensity value Subsequently, an exponential equation of the spectral intensity and slope at different temperatures was established, and the concentration values ​​at different temperatures were predicted based on the regression coefficient weighted correction method to realize the component detection of variable temperature LIBS.

[0102] As can be seen from the above technical solutions, the embodiment of the present invention provides a LIBS spectral modeling method for time-varying temperature of molten steel during RH refining, establishes an exponential equation for spectral intensity and slope at different temperatures, and predicts concentration values ​​at different temperatures through a weighted correction method for regression coefficients. This method is suitable for LIBS composition detection under time-varying temperature conditions of molten steel. The industrial computer integrates a LIBS composition detection algorithm, software, and hardware system for long-distance optical paths to achieve real-time online LIBS composition prediction of molten steel, eliminating the original offline pneumatic sampling, sample preparation, and manual detection steps, saving labor costs, improving smelting detection efficiency, and achieving a high level of system integration and intelligence.

[0103] The above description is only a preferred embodiment of the present invention and an illustration of the technical principles used. It is not intended to limit the scope of the invention to be protected, but merely represents a preferred embodiment of the present invention. Those skilled in the art should understand that the scope of the invention involved in the present invention is not limited to the technical solutions formed by a specific combination of the above-mentioned technical features, but should also cover other technical solutions formed by any combination of the above-mentioned technical features or their equivalent features without departing from the inventive concept. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making any creative work shall fall within the scope of protection of the present invention.

Claims

1. A LIBS spectral modeling method for temperature-varying molten steel during RH refining, characterized in that: The method comprises the following steps: Step S1, obtaining LIBS spectrum data of molten steel in an RH refining furnace; the spectrum data includes spectrum intensity matrix data of different spectral line wavelengths of several elements; Step S2, selecting n modeling analysis line data of a certain element from the spectral intensity matrix data; Step S3, for the n modeling analysis lines of a certain element, respectively calculate the mean and variance of the spectral intensity of each analysis line N times of excitation; Step S4: compare n analysis lines and select the spectral intensity value with the minimum variance of N excitations The corresponding analysis line is used as the starting analysis line λ′; Step S5: Taking the starting analysis line λ′ as the starting point, and drawing the slope curves |K1|, |K2|, ..., |K at room temperature T0 with the remaining (n-1) analysis lines. (n-1) |; corresponding to the high temperature T slope curve |K′1|,|K′2|,...,|K′ (n-1) |; The spectral intensity values ​​of n analysis lines at room temperature T0 are I1, I2, ..., I n , corresponding to the high temperature T spectrum intensity values ​​I′1, I′2, ..., I′ n ; Step S6: According to the blackbody radiation spectrum distribution of Planck's law, when the temperature rises, the absolute value of the slope and the spectral intensity increases exponentially, and the exponential relationship between the slope and the temperature and the spectral intensity are obtained; and further, based on the spectral intensity values ​​corresponding to the room temperature T0 and the temperature T of the high-temperature molten steel liquid, the slope sensitivity coefficient a is derived. i And the sensitivity coefficient b of spectral intensity to temperature i ; Step S7, based on the slope sensitivity coefficient a to temperature i And the sensitivity coefficient b of spectral intensity to temperature i , the predicted element concentration C(T) at different temperatures is corrected by weighting the regression coefficient.

2. The method according to claim 1, characterized in that In step S2, the PLSR regression coefficient feature selection method is used, and the non-interference of element spectral lines and the relatively high intensity persistent lines matched by the NIST database are considered. Finally, n spectral lines are selected as analysis lines for modeling.

3. The method according to claim 2, characterized in that When the ICCD spectrometer used to collect spectral data has a wavelength range of 199-936 nm, the corresponding raw spectral data contains 26,937 wavelength spectral lines. When selecting analytical lines from these 26,937 wavelength spectral lines, 150 spectral lines with large absolute values ​​of regression coefficients are first selected from the 26,937 wavelengths using the PLSR regression coefficient feature selection method. Then, the spectral data of these 150 spectral lines are used to find the spectral line peak points using the second-order derivative extreme point method and matched with the NIST database. The threshold for matching the peak point with the element NIST database is set to 0.

05. When the difference between the peak point and the value in the NIST database is less than 0.05, it is assumed to be the analytical line of the element; otherwise, it is not. At the same time, the element interference lines are eliminated, including the spectral lines of Fe, Cr, and Ni metal elements. Finally, n spectral lines with high relative intensity values ​​are selected as analytical lines for modeling.

4. The method according to claim 1, wherein In step S3, the mean value of N excitations of a single analysis line is: The variance of N excitations of a single analysis line is: In formulas (1) and (2), I k is the spectral intensity value of a single analytical line; is the mean spectral intensity of N excitations of a single analytical line; is the spectral intensity variance of N excitations of a single analytical line; (k = 1, 2, ..., N).

5. The method according to claim 4, characterized in that In step S4, the spectral intensity value with the minimum variance for: The corresponding analysis line is: In formula (4), λ′ is the minimum variance spectral intensity value Corresponding analysis line; (i=1,2,...,n).

6. The method according to claim 5, characterized in that In step S6, the blackbody radiation spectrum distribution formula is as follows: In formula (5), I(v,T) is the radiation intensity at frequency v; h is Planck's constant; c is the speed of light; v is the frequency of radiation; T is the temperature; The exponential relationship between the slope and temperature is: In formula (6), |K i (T0)| is the slope at room temperature T0; a i is a constant term, which is the sensitivity of the slope to temperature; According to the corresponding slope values ​​at room temperature T0 and high temperature molten steel temperature T, the sensitivity coefficient value a of the slope to temperature is derived. i : The exponential relationship between spectral intensity and temperature is: In formula (8), I i+1 (T0) is the spectral intensity at room temperature T0; b i is a constant term, which represents the sensitivity of spectral intensity to temperature; According to the corresponding spectral intensity values ​​at room temperature T0 and high temperature molten steel temperature T, the sensitivity coefficient value b of spectral intensity to temperature is derived. i :

7. The method according to claim 6, characterized in that Calculate the sensitivity coefficient b of spectral intensity and slope to temperature i and a i When the temperature is set to 25°C, the room temperature T0 is set to 25°C and the high temperature molten steel temperature T is set to 1500°C.

8. The method according to claim 6, characterized in that In step S7, the element concentration prediction value C(T) is calculated as follows: Further optimizing formula (10) yields: In formulas (10) and (11), β i+1 is the regression coefficient (or weight) of the i-th spectral line. The regression coefficient value can be solved by the PLSR algorithm through the multivariate regression equation of the normal temperature spectral intensity matrix independent variable I and the element concentration vector dependent variable C; (i=1,2,...,(n-1)); a i is the sensitivity coefficient of slope to temperature; b i is the sensitivity coefficient of spectral intensity to temperature; is the spectral intensity value with minimum variance for n analysis lines and N excitations; I′ i+1 (T) is the spectral intensity at different high temperatures T; |K′ i (T)| is the slope at different high temperatures T; I i+1 (T0) is the spectral intensity at room temperature T0; |K i (T0)| is the slope at room temperature T0.

Images