Molten steel component online detection device and method

Abstract

The invention discloses a molten steel component online detection device and method. The device comprises a laser, an immersion probe and a spectral analysis assembly, the immersion type probe is used for outputting laser emitted by the laser to the surface of molten steel and outputting a plasma spectrum reflected by the surface of the molten steel to the spectrum analysis assembly; the immersion type probe comprises a shell and a light path assembly installed in the shell. A first input end, a first output end and a second output end are arranged on the shell; the first input end is used for receiving laser emitted by a laser; the light path assembly is used for focusing laser emitted by the laser and outputting the laser to the molten steel surface through the first output end; and the plasma spectrum receiving assembly is used for receiving the plasma spectrum reflected by the molten steel surface through the first output end and outputting the plasma spectrum to the spectrum analysis assembly through the second output end. The invention discloses an immersion type probe, which is combined with an LIBS (laser-induced breakdown spectroscopy) technology, realizes real-time online detection of molten steel components, and has a better detection effect.

Description

Technical Field

[0001] The present invention relates to the technical field of smelting production detection equipment, and more specifically to an online detection device and method for molten steel composition. Background Art

[0002] The detection of molten steel composition is a crucial link in the steel production process and has a direct impact on improving product quality and production efficiency. At present, the method for detecting molten steel composition mainly adopts laboratory offline detection, that is, sampling from molten steel during the production process, and sending the sample to the laboratory for chemical analysis after cooling. Although this method can obtain composition data, it has many shortcomings. Its problems are mainly concentrated in detection lag, high operational complexity and high maintenance cost.

[0003] In response to these problems, laser induced breakdown spectroscopy (LIBS) technology provides a new possibility for online detection of molten steel composition. LIBS technology is an element detection method based on laser excitation of the sample surface to generate plasma and analyze the plasma spectrum. It has the advantages of rapid, non-contact, and real-time detection. However, although LIBS technology has significant advantages over traditional detection methods, its application in actual steel production still faces many challenges.

[0004] The primary problem is poor adaptability to the detection environment. The molten steel environment is complex and harsh. Factors such as high temperature, high dust, and strong electromagnetic interference make LIBS equipment easily damaged or contaminated, affecting the continuity and reliability of online detection. In addition, LIBS technology also faces serious signal interference problems in online detection of molten steel composition. High temperature radiation, strong background light, and dust particles in the molten steel environment may interfere with the LIBS spectral signal, resulting in a significant increase in signal noise, weakening the signal-to-noise ratio, and thus reducing the accuracy of the detection results.

[0005] Therefore, how to improve the analysis capabilities of complex spectral data, reduce the impact of environmental factors on detection accuracy, and improve the quality control level and production efficiency of steel production are issues that technical personnel in this field urgently need to solve. Summary of the invention

[0006] In view of this, the present invention provides an online detection device and method for molten steel composition, which, combined with LIBS technology, realizes real-time online detection of molten steel composition and has better detection effect.

[0007] In order to achieve the above object, the present invention adopts the following technical solution:

[0008] An online detection device for molten steel composition comprises a laser, an immersion probe and a spectrum analysis component; the immersion probe is used to output the laser emitted by the laser to the surface of the molten steel, and output the plasma spectrum reflected by the surface of the molten steel to the spectrum analysis component.

[0009] The immersion probe comprises a shell and an optical path component installed inside the shell; the shell is provided with a first output end and a second output end; the optical path component is provided with a first input end, and the first input end is used to receive the laser emitted by the laser.

[0010] The optical path component is used to focus the laser light emitted by the laser and output it to the surface of the molten steel through the first output end; it is used to receive the plasma spectrum reflected by the surface of the molten steel through the first output end, and output the plasma spectrum to the spectral analysis component through the second output end.

[0011] Preferably, the housing is further provided with a second input end and a third output end; the second input end is used for filling with protective gas to protect the optical path component; the third output end is used for exhausting the protective gas.

[0012] Preferably, the optical path component includes a dichroic mirror, a biconvex lens and a single convex lens; the laser passes through the dichroic mirror and is focused by the biconvex lens and output through the first output end; the reflected plasma spectrum passes through the biconvex lens and is reflected by the dichroic mirror, passes through the single convex lens and is output through the second output end.

[0013] Preferably, the material of the shell is austenitic chromium-nickel stainless steel.

[0014] Preferably, the first output end is a conical tube, and the contraction opening of the conical tube is the output port.

[0015] Preferably, it further comprises a lifting device, which is connected to the immersion probe and is used to control the horizontal height of the immersion probe.

[0016] The lifting device includes a ball screw, a coupling, a motor and an L-shaped plate; one plate surface of the L-shaped plate is fixed to the immersion probe, and the other plate surface is fixed to the ball screw; the motor is connected to the ball screw through the coupling; when the motor rotates, the rotational motion is transmitted to the ball screw through the coupling, and the ball screw performs linear motion, thereby driving the L-shaped plate to control the horizontal height of the immersion probe.

[0017] A method for online detection of molten steel composition, comprising the following steps:

[0018] S1: Collecting a plasma spectrum carrying information on the composition of molten steel, splitting the plasma spectrum and performing intensity analysis to obtain the intensity of each wavelength in the spectrum as original spectrum data.

[0019] S2: After preprocessing the raw spectral data, quantitative analysis is performed using a least squares regression model to obtain the element concentration of each component.

[0020] Preferably, the preprocessing includes noise removal, baseline correction and spectral normalization.

[0021] Preferably, the pretreatment step comprises:

[0022] S21: Dynamic weighted sliding average filtering is used to smooth the spectral intensity sequence and eliminate high-frequency noise.

[0023] S22: Correct the baseline by a polynomial fitting method to obtain a corrected spectral intensity sequence; wherein the fitting formula is:

[0024]

[0025] Among them, B(x) is the background signal baseline fitting curve, x represents the wavelength of the spectrum, j represents the polynomial order, which increases from 0 to n, and c j (x) are the polynomial coefficients.

[0026] S23: Normalize the corrected spectral intensity sequence:

[0027]

[0028] Where R(x) is the local intensity ratio adjustment factor, S 校正 (x) is the corrected spectral intensity, S 归一化 (x) is the normalized result.

[0029] Preferably, in S2, the spectral data is quantitatively analyzed by a least squares regression model, the steps comprising:

[0030] S241: The preprocessed spectral data is mapped to the latent variable space through the projection matrix.

[0031] S242: Perform adaptive weight calculation according to the spectral data to obtain an adaptive weight matrix.

[0032] S243: Introduce the adaptive weight matrix into the latent variable space to calculate the element concentration. The prediction formula of the element concentration is:

[0033]

[0034] in, is the predicted element concentration, T is the latent variable, W is the adaptive weight, X input is the preprocessed spectral data, P is the projection matrix of the least squares regression model, and C is the regression coefficient matrix.

[0035] It can be seen from the above technical solutions that, compared with the prior art, the present invention discloses an online detection device and method for molten steel composition, which realizes real-time online detection of molten steel composition by combining LIBS technology with an immersion probe; the immersion probe of the present invention solves environmental interference, improves detection accuracy and data stability. In addition, combined with argon protection, it reduces the loss of the probe due to high temperature and increases the service life of the equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0037] Figure 1 A schematic structural diagram of an online detection device for molten steel composition provided by the present invention;

[0038] Figure 2 A schematic flow chart of an online method for detecting molten steel composition provided by the present invention. DETAILED DESCRIPTION

[0039] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0040] Example 1

[0041] like Figure 1 The embodiment of the present invention discloses an online detection device for molten steel composition, which is composed of the following parts:

[0042] Lasers are used to emit high-energy laser beams. High-energy laser beams can stimulate the surface of molten steel to form plasma. Different molten steel components can produce different plasma spectra. The composition of molten steel can be obtained by analyzing the plasma spectrum. The selection of lasers meets the requirements of high stability and ensures the stable generation of plasma.

[0043] The spectrum analysis component is used to realize the analysis of plasma spectrum, which may include a spectrometer and a computer.

[0044] Immersed probe used to guide the laser to the steel surface and provide a channel for collecting plasma spectra.

[0045] In this embodiment, the laser emitted by the laser is first irradiated onto the surface of the molten steel under the guidance of the immersion probe to form a reflected plasma spectrum. The reflected plasma spectrum is received by the immersion probe again, and the plasma spectrum is output to the corresponding position by changing the propagation path so that it can be acquired by the spectral analysis component.

[0046] In order to further implement the above technical solution, the transmission of the plasma spectrum to the spectrum analysis component can also be achieved through the corresponding optical fiber structure. The reflected plasma spectrum is transmitted to the optical fiber inlet under the action of the immersion probe, and the optical fiber outlet is connected to the spectrum analysis component.

[0047] In order to further implement the above technical solution, the immersion probe includes a housing 1 and an optical path component installed inside the housing 1 , and the housing 1 is provided with a first output terminal 2 and a second output terminal 3 .

[0048] The optical path component is provided with a first input terminal 4, and the first input terminal 4 is used to receive the laser emitted by the laser; the optical path component is used to focus the laser emitted by the laser and output it to the surface of the molten steel through the first output terminal 2; it is used to receive the plasma spectrum reflected by the surface of the molten steel through the first output terminal 2, and construct the plasma spectrum and output it to the spectrum analysis component through the second output terminal 3.

[0049] Among them, the optical path components are responsible for the transmission and focusing of laser and plasma spectra, which mainly include dichroic mirrors, double convex lenses and single convex lenses.

[0050] The dichroic mirror is placed at an angle with the preferred angle of 45°. It has the characteristics of high transmittance to light of specific wavelengths and high reflection to light of other wavelengths. It can realize the separation and integration of laser and spectral signals. Under the action of the dichroic mirror, it can transmit the laser emitted by the laser and reflect the plasma spectrum. Under the design of the tilt angle, the reflection angle can be accurately controlled, and the propagation path of the plasma spectrum can be controlled.

[0051] The main function of the biconvex lens is focusing. It can focus the light transmitted through the dichroic mirror onto the surface of the molten steel; at the same time, it can also focus the reflected plasma spectrum onto the dichroic mirror.

[0052] The dichroic mirror reflects the focused plasma spectrum to the single convex lens, which then focuses the spectrum to the second output end in one step; the second output end may be an optical fiber port.

[0053] In addition, in order to allow the laser emitted by the laser to be accurately incident, a total reflection mirror may be provided to control the path of the laser and reflect the laser path to the dichroic mirror.

[0054] In this embodiment, the shell mainly includes a flange sleeve, and various mounting frames are arranged inside the flange sleeve for mounting various optical components, such as a double convex lens mounting frame 5, a dichroic mirror mounting frame 6, a single convex lens mounting frame 7 and an optical fiber mounting frame.

[0055] The housing 1 is made of austenitic chromium-nickel stainless steel. This material has excellent corrosion resistance and good mechanical properties, can effectively isolate dust from the external environment, and protect the internal precision optical components from damage. The flange sleeve effectively prevents molten steel from splashing into the probe, ensuring the cleanliness of the optical path and the accurate transmission of the spectral signal.

[0056] Specifically, the shell adopts a double sleeve design, and the optical component is fixed separately by the upper sleeve 11. Under the action of the lower sleeve 12, the optical component is effectively isolated from the high temperature environment of the molten steel below it. In addition, in addition to its supporting function, the lower sleeve 12 can also stabilize the gas environment.

[0057] Furthermore, the first output end 2 is a conical tube, which is the part of the immersion probe that is in direct contact with the molten steel and is made of alumina ceramic material. Alumina ceramics have excellent high temperature resistance, corrosion resistance and high strength characteristics, and can work stably for a long time in a high temperature environment of molten steel. The conical design of the conical tube helps to reduce the adhesion of molten steel and avoid the accumulation of slag and impurities on the surface. In addition, the low thermal conductivity of the alumina material also reduces the impact of heat conduction on the internal optical components, effectively protecting the long-term stable operation of the optical components.

[0058] In order to further implement the above technical solution, the housing 1 is further provided with a second input terminal 8 and a third output terminal 9; the second input terminal 8 is used to fill with protective gas to protect the optical path components; the third output terminal 9 is used to discharge the protective gas.

[0059] In this embodiment, the interior of the immersion probe can be filled with a protective gas, preferably argon, through the second input terminal 8. The second input terminal 8 serves as an air inlet, which is arranged at the top of the shell, and correspondingly, the third output terminal 9 is an air outlet, which can be arranged in the lower area of ​​the entire probe. By filling the air inlet with high-purity argon, the air inside the probe can be discharged to form an argon protection environment. Since the density of argon is greater than that of air, argon can effectively replace the air in the probe to prevent oxygen, nitrogen, etc. in the environment from reacting with the excited plasma. Another important function of argon protection is to cool the probe. In the working environment of high-temperature molten steel, the probe is immersed in the molten steel for detection, which is easily affected by the high temperature and causes the internal temperature to rise. The continuous filling of argon can form a stable airflow inside the probe, which takes away the heat on the probe surface and internal optical components by convection heat dissipation, thereby effectively reducing the overall temperature of the probe.

[0060] Among them, in the double-layer sleeve structure, the gas outlet is set on the side wall of the lower sleeve. A one-way valve can be used, and the clock is kept in an open state. The atmosphere environment is controlled by adjusting the argon flow rate. The argon supply system is used to maintain the appropriate internal pressure, which causes the gas to flow out of the gas outlet through the pressure difference, forming an effective protective gas barrier.

[0061] In order to further implement the above technical solution, a lifting device 10 is also included, which is connected to the immersion probe and is used to control the horizontal height of the immersion probe. During the detection process, the lifting device 10 controls the probe to be steadily lowered into the molten steel for detection; after the detection, the device smoothly raises the probe, effectively avoiding the loss of the probe due to long-term exposure to high-temperature molten steel, and facilitating subsequent maintenance of the probe.

[0062] Example 2

[0063] like Figure 2 Based on the same inventive concept, the embodiment of the present invention discloses an online detection method for molten steel composition, comprising the following steps:

[0064] S1: Collect the plasma spectrum carrying the composition information of the molten steel, split the plasma spectrum and perform intensity analysis to obtain the intensity of each wavelength in the spectrum as the original spectrum data.

[0065] S2: After preprocessing the raw spectral data, quantitative analysis is performed using the least squares regression model to obtain the element concentration of each component.

[0066] In order to further implement the above technical solution, S2 specifically includes:

[0067] S21: Remove the high-frequency interference signal in the spectrum through the noise removal step, use the sliding average filter to smooth the spectrum intensity sequence, and calculate the weighted average of each data point through the sliding window to remove the high-frequency noise and retain the important spectral characteristic peaks. Specifically, the formula of the smoothing filter is as follows:

[0068]

[0069] Where S(i) is the i-th spectral intensity value; N is the number of sampling points in the window; w k is the weight factor.

[0070] In this embodiment, the weight factor can be adaptively adjusted according to the local gradient change, and the adjustment method is:

[0071]

[0072] Among them, k is the data point variable in the sliding window.

[0073] S22: A polynomial fitting method is used to estimate and subtract the background signal so that the corrected spectrum can reflect the true element concentration. An n-order polynomial is used to approximate the background signal and subtract it from the original spectrum to obtain the corrected spectrum intensity.

[0074] Specifically, the polynomial fitting formula is:

[0075]

[0076] Among them, the coefficient c j (x) Dynamically adjust according to changes in the width of adjacent peaks in the spectrum, allowing the baseline to flexibly adapt to complex backgrounds.

[0077] After obtaining the fitted polynomial baseline, the baseline is corrected by the polynomial fitting method. The corrected spectral intensity is: S 校正 (x) = S 平滑 (x)-B(x).

[0078] S23: Use the unit vector normalization method to adjust the corrected spectral data to a uniform scale to eliminate the impact of changes in measurement conditions. After normalization, each spectral data point is converted into a unit vector, allowing different spectra to be compared within the same intensity range, effectively improving data consistency and providing stable input for model analysis.

[0079] Specifically, the normalization process combines the unit vector normalization and the local intensity ratio adjustment process. The normalization formula is:

[0080]

[0081] Among them, R(x) is the local intensity ratio adjustment factor. Through this adjustment factor, the excessive amplification or compression effect that may be introduced in the normalization process can be suppressed.

[0082] S24: Quantitative analysis of spectral data was performed using the least squares regression (PLSR) model. The PLSR model achieves regression analysis of spectral features and element concentrations by reducing the dimensionality of spectral data to a few latent variable spaces. The input matrix consists of spectral data, and the output is the actual value of element concentration. PLSR maps high-dimensional spectral data to latent variables through a projection matrix, and estimates model parameters by minimizing the residual sum of squares, thereby achieving concentration prediction.

[0083] S241: The normalized spectral data S 归一化 (x) is input as the input matrix in the quantitative analysis step. The input matrix is ​​in the form of: X 输入 =[S 归一化 (x1),S 归一化 (x2),…,S 归一化 (x n )] T , where x1,x2,…,x n is the spectral intensity at different wavelengths. Based on the normalized input matrix X input, the PLSR model maps the spectral data to the latent variable space through the projection matrix P. The latent variable T is a low-dimensional representation of the spectral features.

[0084] S242: Calculate the adaptive weights based on the spectral data to obtain an adaptive weight matrix

[0085] S243: Introducing the adaptive weight matrix into the latent variable space to calculate the element concentration; the prediction formula of the element concentration is:

[0086]

[0087] in, is the predicted element concentration, T is the latent variable, W is the adaptive weight, X input is the preprocessed spectral data, P is the projection matrix of the least squares regression model, and C is the regression coefficient matrix.

[0088] First, in the noise removal step, a dynamic weighted sliding average filter is used to smooth the spectral intensity sequence to remove high-frequency noise. Specifically, the weighted average of each data point is calculated through a sliding window, and the weight is adaptively adjusted according to the local gradient change.

[0089]

[0090] Where: S(i) is the i-th spectral intensity value; N is the number of sampling points in the window; w kis the weight factor, M is the width parameter of the sliding window, which is half of the sliding window; S(i+k) represents the value of the i+kth position in the original spectral intensity sequence; k is the relative offset within the window, ranging from -M to M.

[0091] The adaptive adjustment method of the weight factor is expressed as:

[0092]

[0093] Secondly, the baseline is corrected by polynomial fitting method, and the background signal is represented by an adaptive n-order polynomial. A dynamic polynomial fitting formula based on peak width and position is used:

[0094]

[0095] Where B(x) represents the background signal, x represents the wavelength of the spectrum, j represents the polynomial order, which increases from 0 to n; the coefficient c j (x) Dynamically adjust according to the change of adjacent peak width in the spectrum, so that the baseline can flexibly adapt to the complex background. The corrected spectral intensity is:

[0096] S 校正 (x) = S 平滑 (x)-B(x)

[0097] After the baseline correction is completed, the data needs to be further normalized to eliminate the influence of changes in measurement conditions and experimental environment. A method combining unit vector normalization and local intensity ratio adjustment is used.

[0098]

[0099] Wherein, R(x) is a local intensity ratio adjustment factor, through which the over-amplification or compression effect that may be introduced in the normalization process can be suppressed.

[0100] After data preprocessing, the spectral data were quantitatively analyzed using the improved partial least squares regression (PLSR) model. 归一化 (x) is input as the input matrix X in the quantitative analysis step. The input matrix is ​​in the form of: X 输入 =[S 归一化 (x1),S 归一化 (x2),...,S 归一化 (x n )] T , where x1,x2,...,x n is the spectral intensity at different wavelengths.

[0101] Then, based on the normalized input matrix X input, the PLSR model maps the spectral data to the latent variable space through the projection matrix P. The latent variable T is a low-dimensional representation of the spectral features, and the mapping formula is:

[0102] T=X 输入 P

[0103] In the latent variable space after projection, an adaptive weight matrix W is introduced. The weight matrix W is based on the variance change of the spectrum:

[0104]

[0105] The final element concentration prediction formula is:

[0106]

[0107] in, is the predicted element concentration, and C is the regression coefficient matrix.

[0108] In this embodiment, the model accuracy is improved by introducing an adaptive weight matrix, and the adaptive weight matrix W can dynamically adjust the weight according to the variance changes of the spectral data at different wavelengths. : In the spectral data, the signal intensity changes at different wavelengths are different, and these changes can be measured by the variance. Generally speaking, the bands with large variances mean that the spectral data contain more useful information at these wavelengths, which may be more relevant to the target variable (element concentration). The bands with small variances contain less information and may be noise. The adaptive weight matrix W adjusts the weights based on the variance of each wavelength, giving higher weights to wavelength regions with larger variances, and giving lower weights to regions with smaller variances or smaller fluctuations. This adaptive adjustment method can ensure that the model focuses on bands that are useful for prediction, rather than being disturbed by noise or irrelevant information. By adaptively adjusting the weights, the model can give greater influence to important spectral regions, thereby improving the accuracy of the prediction. In addition, by dynamically adjusting the weights according to the variance of spectral features of different wavelengths, the model can better adapt to different types of data and reduce the possibility of overfitting.

[0109] The present invention ensures the accuracy and reliability of molten steel component detection by performing a series of preprocessing and quantitative calculation on spectral data, and provides reliable data support for real-time control of the production process.

[0110] The optimization of material selection and structural design of the immersion probe effectively extends the service life of the equipment. The argon protection and optical path design inside the probe effectively reduce the impact of environmental interference, oxygen reaction and high temperature on the equipment. The lifting mechanism of the probe simplifies the detection operation and subsequent maintenance process, further improving the overall operating efficiency and durability of the equipment.

[0111] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0112] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An online detection device for molten steel composition, characterized in that: It comprises a laser, an immersion probe and a spectrum analysis component; the immersion probe is used to output the laser emitted by the laser to the surface of molten steel, and output the plasma spectrum reflected by the surface of molten steel to the spectrum analysis component; The immersion probe comprises a housing and an optical path component installed inside the housing; the housing is provided with a first output end and a second output end; The optical path component is provided with a first input end, and the first input end is used to receive the laser emitted by the laser; The optical path component is used to focus the laser light emitted by the laser and output it to the surface of the molten steel through the first output end; it is used to receive the plasma spectrum reflected by the surface of the molten steel through the first output end, and output the plasma spectrum to the spectral analysis component through the second output end.

2. The on-line detection device for molten steel composition according to claim 1, characterized in that: The housing is also provided with a second input end and a third output end; the second input end is used for filling with protective gas to protect the optical path component; the third output end is used for exhausting the protective gas.

3. The on-line detection device for molten steel composition according to claim 1, characterized in that: The optical path components include a dichroic mirror, a biconvex lens and a single convex lens; The laser passes through the dichroic mirror and is focused by the biconvex lens, and is output through the first output end; The reflected plasma spectrum passes through the double convex lens and then, under the reflection of the dichroic mirror, passes through the single convex lens and is output through the second output end.

4. The on-line detection device for molten steel composition according to claim 1, characterized in that: The material of the shell is austenitic chromium-nickel stainless steel.

5. The on-line detection device for molten steel composition according to claim 1, characterized in that: The first output end is a conical tube, and the contraction opening of the conical tube is the output opening.

6. The on-line detection device for molten steel composition according to claim 1, characterized in that: It also includes a lifting device, which is connected to the immersion probe and is used to control the horizontal height of the immersion probe; The lifting device includes a ball screw, a coupling, a motor and an L-shaped plate; one plate surface of the L-shaped plate is fixed to the immersion probe, and the other plate surface is fixed to the ball screw; the motor is connected to the ball screw through the coupling; when the motor rotates, the rotational motion is transmitted to the ball screw through the coupling, and the ball screw performs linear motion, thereby driving the L-shaped plate to control the horizontal height of the immersion probe.

7. A method for online detection of molten steel composition, characterized in that: The on-line detection device for molten steel composition according to any one of claims 1 to 7 is adopted, and the steps include: S1: Collecting a plasma spectrum carrying information on the composition of molten steel, splitting the plasma spectrum and performing intensity analysis to obtain the intensity of each wavelength in the spectrum as original spectrum data; S2: After preprocessing the raw spectral data, quantitative analysis is performed using a least squares regression model to obtain the element concentration of each component.

8. The method for online detection of molten steel composition according to claim 7, characterized in that: The preprocessing includes noise removal, baseline correction and spectral normalization.

9. The method for online detection of molten steel composition according to claim 7, characterized in that: The pre-processing steps include: S21: Dynamic weighted sliding average filtering is used to smooth the spectral intensity sequence and eliminate high-frequency noise; S22: Correct the baseline by a polynomial fitting method to obtain a corrected spectral intensity sequence; wherein the fitting formula is: Among them, B(x) is the background signal baseline fitting curve, x represents the wavelength of the spectrum, j represents the polynomial order, which increases from 0 to n, and c j (x) are the polynomial coefficients. S23: Normalize the corrected spectral intensity sequence: Where R(x) is the local intensity ratio adjustment factor, S 校正 (x) is the corrected spectral intensity, S 归一化 (x) is the normalized result.

10. A method for online detection of molten steel composition according to claim 7 or 9, characterized in that: In S2, the spectral data is quantitatively analyzed by a least squares regression model, and the steps include: S241: The preprocessed spectral data is mapped to the latent variable space through the projection matrix; S242: performing adaptive weight calculation according to the spectral data to obtain an adaptive weight matrix; S243: Introducing the adaptive weight matrix into the latent variable space to calculate the element concentration; the prediction formula of the element concentration is: in, is the predicted element concentration, T is the latent variable, W is the adaptive weight, X input is the preprocessed spectral data, P is the projection matrix of the least squares regression model, and C is the regression coefficient matrix.

Images