A method and system for monitoring the grinding of magnesium oxide

Abstract

The invention discloses a method and system for monitoring the grinding of magnesium oxide, and specifically relates to the technical field of grinding monitoring. The invention first collects a reference spectrum generated by light reflection from the surface of a reference substrate with a target thickness; during the grinding process, a plurality of optical sensors are used to capture the reflection spectrum of the surface of the magnesium oxide grinding substrate at multiple angles; then, the image dust influence coefficient of each optical sensor is calculated according to the K value and the spatiotemporal concentration distribution of dust; an angle correction factor and the image dust influence coefficient are determined to comprehensively allocate the weight coefficient of the reflection spectrum obtained by each optical sensor; the reflection spectra obtained by each optical sensor are fused and output according to the weight coefficient of the reflection spectrum obtained by each optical sensor; finally, the real-time spectrum is matched with a historical spectrum library; the target thickness of each area is compared to determine the rate deviation; the grinding pressure of different areas is dynamically adjusted through a pressure control system, and the grinding endpoint time of all areas is made to be close to the same.

Description

Technical Field

[0001] The present invention relates to the technical field of grinding monitoring, and more particularly to a method and system for monitoring the grinding of magnesium oxide. Background Art

[0002] Magnesium oxide (MgO) is a commonly used inorganic compound used in a wide range of applications, including building materials, agriculture, medicine, chemicals, and environmental protection. Grinding is a crucial step in the magnesium oxide production process. It involves physically refining magnesium oxide powder to increase its surface activity and reactivity, thereby improving its performance in various applications.

[0003] Chemical mechanical polishing (CMP) is a well-established planarization method. This planarization method generally requires the substrate to be placed on a carrier or polishing head. The exposed surface of the substrate is typically positioned relative to a rotating disk or belt polishing pad. The polishing pad can be a standard polishing pad or a fixed abrasive polishing pad. Standard polishing pads have a permanently roughened surface, while fixed abrasive polishing pads have abrasive particles held in a containment medium. The carrier head provides a controlled load on the substrate to push it toward the polishing pad. A slurry is typically supplied to the polishing pad surface. The slurry contains at least one chemical reactant for a standard polishing pad and abrasive particles.

[0004] In the prior art, the grinding process of magnesium oxide easily generates dust and slurry, which affects the accuracy of the reflectance spectrum and causes errors in subsequent operations based on the reflectance spectrum. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art, embodiments of the present invention provide a method and system for monitoring the grinding of magnesium oxide to solve the problems raised in the above-mentioned background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A method for monitoring the grinding of magnesium oxide comprises the following steps:

[0008] Collect a reference spectrum generated by light reflection from the surface of a reference substrate with a target thickness; during the polishing process, multiple optical sensors are used to capture the reflectance spectrum of the magnesium oxide polishing substrate surface at multiple angles;

[0009] Calculate the dust impact coefficient of each optical sensor based on the K value and the spatiotemporal distribution of dust concentration; determine the angle correction factor and the dust impact coefficient to comprehensively allocate the reflection spectrum weight coefficient obtained by each optical sensor; and fuse the reflection spectra obtained by each optical sensor according to the reflection spectrum weight coefficient obtained by each optical sensor.

[0010] The real-time spectrum is matched with the historical spectrum library to generate thickness indicators for each area, which are real-time thickness estimates. The target thickness of each area is compared to determine the rate deviation. The carrier head zone pressure control system dynamically adjusts the grinding pressure in different areas to ensure that the grinding endpoint time of all areas is close to the same.

[0011] The K value determines the capture range of the optical sensor, and compares the capture range of the optical sensor with the predicted dust movement trajectory; if the optical sensor is blocked, K=1, otherwise, K=0.

[0012] In a preferred embodiment, the predicted dust movement trajectory is achieved through an Euler-Lagrange coupling model, including the following steps: first, a three-dimensional geometric calculation domain is constructed based on the CAD model of the grinding equipment, and the airflow field is characterized by the Navier-Stokes equation combined with the turbulence model; and the force and movement trajectory of the dust particles are calculated through the drag model and the lift model.

[0013] In a preferred embodiment, the screen dust impact coefficient of each optical sensor is calculated using a logistic regression formula according to the K value and the spatiotemporal concentration of dust.

[0014] In a preferred embodiment, the spatiotemporal dust concentration represents the spatial dust concentration at different times.

[0015] In a preferred embodiment, the reflection spectrum weight coefficients obtained by each optical sensor are comprehensively calculated based on the angle correction factor and the screen dust influence coefficient.

[0016] In a preferred embodiment, the angle correction factor is determined by the following steps: first, a theoretical light intensity and polarization relationship is established based on the Fresnel reflection model; then, the actual angle of the sensor layout is calibrated by laser, and dynamic interference is compensated in combination with the dust scattering coefficient and the optical path length; finally, the ratio of the measured reflectivity to the theoretical reflectivity is used as the angle correction factor.

[0017] In a preferred embodiment, the reflection spectra acquired by each optical sensor are fused and output according to the reflection spectrum weight coefficients acquired by each optical sensor.

[0018] In a preferred embodiment, the historical spectral library is constructed by: first collecting standard reflectance spectra of magnesium oxide at different thicknesses; then extracting characteristic peak positions, intensity ratios and full spectrum patterns as spectral features; and finally generating a low-dimensional feature library through principal component analysis.

[0019] In a preferred embodiment, the system includes the following modules: a data acquisition module, a data fusion module, and an automatic adjustment module;

[0020] The data acquisition module acquires multi-dimensional dynamic data of the grinding process in real time, providing raw input for subsequent analysis;

[0021] The data fusion module is used to comprehensively allocate the reflection spectrum weight coefficient obtained by each optical sensor according to the angle correction factor and the screen dust influence coefficient; and fuse the reflection spectra obtained by each optical sensor according to the reflection spectrum weight coefficient obtained by each optical sensor and output it;

[0022] The automatic adjustment module dynamically controls the grinding parameters based on the fused data to ensure that the grinding endpoint time of each area is consistent; the real-time reflection spectrum is matched with the historical spectrum library, and the real-time thickness estimation value of each area is generated through principal component analysis; the target thickness is compared with the real-time thickness pointer to calculate the grinding rate deviation of each area; and differentiated grinding pressure is applied to different areas through the carrier head partition pressure control system.

[0023] In a preferred embodiment, the data fusion module includes a dust trajectory calculation module and an image dust impact calculation module;

[0024] The dust trajectory calculation module predicts the dust diffusion trajectory during the grinding process, determines the optical sensor capture range interval, and compares the optical sensor capture range interval with the predicted dust movement trajectory; if the optical sensor is blocked, K=1, otherwise, K=0;

[0025] The image dust impact calculation module calculates the image dust impact coefficient of each optical sensor according to the K value and the spatiotemporal concentration distribution of dust.

[0026] Technical effects and advantages of the present invention:

[0027] The present invention first collects a reference spectrum generated by light reflection from the surface of a reference substrate with a target thickness; during the grinding process, multiple optical sensors are used to capture the reflection spectrum of the magnesium oxide grinding substrate surface at multiple angles; the optical sensor capture range interval is determined, and the optical sensor capture range interval is compared with the predicted dust movement trajectory; if the optical sensor is blocked, K=1, otherwise K=0.

[0028] Then, the dust impact coefficient of each optical sensor's image is calculated based on the K value and the spatiotemporal concentration distribution of dust; the angle correction factor and the dust impact coefficient are determined to comprehensively allocate the reflection spectrum weight coefficient obtained by each optical sensor; the reflection spectrum obtained by each optical sensor is fused and output according to the reflection spectrum weight coefficient obtained by each optical sensor; the interference of dust and mud is avoided, making the obtained reflection spectrum more accurate; finally, the real-time spectrum is matched with the historical spectrum library to generate a "thickness pointer" for each area; the target thickness of each area is compared to determine the rate deviation; the grinding pressure of different areas is dynamically adjusted through the carrier head partition pressure control system to make the grinding endpoint time of all areas close to the same. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings;

[0030] Figure 1 This is a schematic flow chart of a method for monitoring the grinding of magnesium oxide according to the present invention;

[0031] Figure 2 This is a schematic flow chart of a magnesium oxide grinding monitoring system according to the present invention;

[0032] Figure 3 The figure is a structural diagram of a magnesium oxide grinding monitoring system of the present invention. DETAILED DESCRIPTION

[0033] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. 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 making creative efforts are within the scope of protection of the present invention.

[0034] The present invention first collects a reference spectrum generated by light reflection from the surface of a reference substrate with a target thickness; during the grinding process, multiple optical sensors are used to capture the reflection spectrum of the magnesium oxide grinding substrate surface at multiple angles; the optical sensor capture range interval is determined, and the optical sensor capture range interval is compared with the predicted dust movement trajectory; if the optical sensor is blocked, K=1, otherwise K=0.

[0035] Then, the dust impact coefficient of each optical sensor's image is calculated based on the K value and the spatiotemporal concentration distribution of dust. The angle correction factor and the dust impact coefficient are determined to comprehensively allocate the reflection spectrum weight coefficient obtained by each optical sensor. The reflection spectrum obtained by each optical sensor is fused and output according to the reflection spectrum weight coefficient obtained by each optical sensor. Finally, the real-time spectrum is matched with the historical spectrum library to generate a "thickness pointer" for each area. The target thickness of each area is compared to determine the rate deviation. The grinding pressure of different areas is dynamically adjusted through the carrier head partition pressure control system to make the grinding endpoint time of all areas close to the same.

[0036] Example 1

[0037] The present invention provides a method for monitoring the grinding of magnesium oxide, such as Figure 1 As shown, the following steps are included:

[0038] Collect a reference spectrum generated by light reflection from the surface of a reference substrate with a target thickness; during the polishing process, multiple optical sensors are used to capture the reflectance spectrum of the magnesium oxide polishing substrate surface at multiple angles;

[0039] Calculate the dust impact coefficient of each optical sensor based on the K value and the spatiotemporal distribution of dust concentration; determine the angle correction factor and the dust impact coefficient to comprehensively allocate the reflection spectrum weight coefficient obtained by each optical sensor; and fuse the reflection spectra obtained by each optical sensor according to the reflection spectrum weight coefficient obtained by each optical sensor.

[0040] Match the real-time spectrum with the historical spectrum library to generate a "thickness pointer" for each area; compare the target thickness of each area to determine the rate deviation; dynamically adjust the grinding pressure in different areas through the carrier head partition pressure control system to ensure that the grinding endpoint time of all areas is close to the same;

[0041] The K value determines the capture range of the optical sensor, and compares the capture range of the optical sensor with the predicted dust movement trajectory; if the optical sensor is blocked, K=1, otherwise, K=0.

[0042] specific;

[0043] Select the magnesium oxide to be ground, ensuring its initial thickness is greater than the target thickness. Using a white light interferometer or other spectral measurement device, collect reflected white light spectra during the grinding process at fixed time intervals, such as 0.1 second intervals or the grinding platform rotation period. Each spectral data set contains the reflected intensity distribution within a wavelength range (e.g., 400-800 nm). A reference spectrum generated by the white light reflection is collected; this reference spectrum may be single or multiple.

[0044] During the magnesium oxide grinding process, magnesium oxide, magnesium powder, and other impurities are generated. A flow of air (such as nitrogen or dry air) is injected through the delivery nozzles onto the upper surface of the optical head or the lower surface of the polishing pad window to prevent chemical reactions with the magnesium oxide. Simultaneous suction is achieved through the vacuum nozzles to create a laminar airflow pattern, preventing turbulence that can cause contaminants to adhere. This dust, during its movement, can obstruct the slurry, cause slurry splashing, and cause residual bubbles. This can affect the real-time reflectance spectrum acquired by the optical sensor, leading to errors in the acquired reflectance spectrum. Multiple optical sensors are installed to capture reflectance spectra from multiple angles.

[0045] First, the airflow direction is determined. Based on the airflow direction and the grinding position, the dust trajectory is predicted using the Euler-Lagrangian coupling model. Specifically, the Euler-Lagrangian coupling model (ELM) separates the continuous phase (airflow) from the discrete phase (dust particles) and is suitable for dynamic simulation of gas-solid two-phase flow in grinding scenarios. The Euler framework primarily describes the Navier-Stokes equations for the airflow field, accounting for macroscopic flow characteristics such as turbulence and pressure gradients. The Lagrangian framework primarily tracks the forces and motion trajectories of individual particles and is suitable for the discrete characteristics of dust.

[0046] The computational domain is constructed based on the CAD model of the grinding equipment, with a focus on meshing the grinding area and the dust diffusion path (the boundary layer mesh thickness is recommended to be 1 / 3 of the minimum dust particle size).

[0047] First, based on the characteristics of the grinding scene, a computational domain and a 3D geometric model are constructed based on the CAD model of the grinding equipment. Boundary layer mesh encryption technology is used to focus on meshing the grinding area and the dust diffusion path (the boundary layer mesh thickness is recommended to be 1 / 3 of the minimum dust particle size). Then, the Navier-Stokes equations are solved: ; The symbols in the formula are interpreted as follows: is the component of the velocity vector in the i-th Cartesian direction (i=1,2,3 corresponds to the x, y, z axis respectively); is the component of the velocity vector in the jth direction. Since the index j is repeated, it follows the Einstein summation rule, that is, the summation is automatically performed for j = 1 to j = 3; t is time; and Represents the i-th or j-th spatial coordinate respectively; and is the partial derivative operator with respect to time or space coordinates, used to describe local unsteady changes and self-transport along the flow; ρ is the fluid density; p is the absolute static pressure; ν = μ⁄ρ is the kinematic viscosity, where μ is the dynamic viscosity; is the velocity component The Laplace operator reflects the viscous diffusion effect; It is the volume force source term generated by particle-gas coupling, and its magnitude usually depends on the particle diameter, relative velocity and drag coefficient, and is used to quantify the additional resistance or traction of particles on the airflow. and They represent local acceleration and convective acceleration respectively. The right side represents pressure gradient force, viscous diffusion force and particle drag force, respectively, which realizes a complete description of the momentum balance of sparse dust-laden flow field.

[0048] The k-ε SST or LES turbulence model is combined to characterize airflow pulsations. The Rosin-Rammler particle size distribution (typical median size 20 μm) and physical properties of the dust are then defined. The Schiller-Naumann drag model, Saffman lift model, and wall impact model are used to calculate the spatiotemporal variations of the forces acting on the particles. A bidirectional coupling strategy is employed to update the momentum exchange term every 10 fluid time steps. The numerical solution utilizes the PISO algorithm and adaptive step-size control, leveraging GPU parallel acceleration to reduce computation time to within 24 hours. Finally, streamline tracing and comparison with experimental data from a laser particle size analyzer reveal the dust diffusion patterns.

[0049] Finally, the particle paths were integrated using Stream Tracer in Paraview, and the particle position time series was exported (CSV format). The trajectory was visualized using Python Matplotlib dynamic plotting.

[0050] The capture range of the optical sensor is determined according to the installation position of the optical sensor. The capture range of the optical sensor is compared with the predicted dust movement trajectory, and a timestamp is added. If the image captured by the optical sensor at the same timestamp coincides with the predicted dust movement trajectory, K=1, otherwise, K=0.

[0051] The explosion-proof laser dust monitor is installed and timestamped to obtain the spatiotemporal dust concentration. The spatiotemporal dust concentration represents the spatial dust concentration at different times. The dust impact coefficient of each optical sensor is calculated based on the K value and the spatiotemporal dust concentration. Specifically, the logistic regression formula is used to calculate the dust impact coefficient of each optical sensor. The formula is as follows: ;in represents the dust impact coefficient of the i-th optical sensor; fb represents the spatiotemporal concentration of dust; the greater the spatiotemporal concentration of dust, the greater the dust impact coefficient of the sensor image, and vice versa; e represents the natural base; α and β represent the logistic regression coefficients of K value and spatiotemporal concentration of dust, respectively, and both are greater than zero.

[0052] The reflection spectrum with low dust impact coefficient on the screen has higher availability and authenticity; the dust spatiotemporal concentration represents the spatial dust concentration at different times.

[0053] During the magnesium oxide grinding process, optical sensors capture the reflected spectrum at multiple angles to compensate for dust interference. However, differences in the geometric relationship between the incident and reflected angles at different sensor locations can cause spectral signal distortion. Therefore, an angle correction factor is introduced to quantify the degree of spectral distortion. The angle correction factor is determined based on the relationship between the incident and reflected angles.

[0054] The core of the angle correction factor is to quantify the impact of the difference between the incident angle (θ1) and the reflection angle (θ2) on the spectral signal through geometric optical relationships. The specific steps include: first, establishing a theoretical relationship between light intensity and polarization based on the Fresnel reflection model, using a laser-assisted calibration device, and determining the actual angle of the sensor layout through triangulation. The vertical height h and horizontal offset distance d between the sensor installation position and the grinding plane are recorded, and the theoretical angle is calculated using trigonometric functions: ; Secondly, the sensor layout angle is obtained through laser calibration and actual measurement, and the angle correction factor is derived, which is defined as the ratio of the measured reflectivity to the theoretical Fresnel reflectivity. At the same time, the dust scattering coefficient and optical path length are introduced to compensate for dynamic dust interference; the experimental phase requires establishing a dust-free reference angle correction factor in a clean environment, and then combining the real-time dust concentration data to dynamically optimize the model. Finally, the accuracy of the angle correction factor is ensured through cross-validation (such as using the vertical incidence sensor as a reference) and adaptive algorithms (such as Kalman filtering).

[0055] The reflection spectrum weight coefficient obtained by each optical sensor is calculated comprehensively based on the angle correction factor and the screen dust influence coefficient. Specifically, the reflection spectrum weight coefficient is calculated by the weighted summation formula, which is as follows: ; represents the i-th reflection spectrum weight coefficient; yz represents the angle correction factor; λ and δ are the weight coefficients of the angle correction factor and the screen dust influence coefficient, respectively.

[0056] The reflection spectra obtained by each optical sensor are fused and output according to the reflection spectrum weight coefficients obtained by each optical sensor.

[0057] The reference spectrum is stored in a historical spectral library; the historical spectral library pre-constructs a standard spectral feature library of magnesium oxide solids at different thicknesses through experiments or simulations; the historical spectral library includes the following dimensions: characteristic peak position (such as the linear / nonlinear relationship between the offset of a specific wavelength peak and thickness); intensity ratio (such as peak height ratio, integrated area ratio); full spectrum mode (low-dimensional features are extracted through principal component analysis / PCA or deep learning models).

[0058] The collected reflectance spectra are preprocessed, the real-time signals are subjected to Savitzky-Golay filtering to eliminate noise, and the historical spectrum library is aligned through wavelength calibration; polynomial fitting or adaptive iterative algorithm is used to remove background interference.

[0059] The real-time reflectance spectrum within a sliding time window is used to perform similarity calculations (such as Pearson correlation coefficient and dynamic time warping / DTW) with the historical library to screen out the best matching reference spectrum.

[0060] Then, the thickness pointer is calculated and the single peak is located. If there is a mapping relationship between the characteristic peak position (λ) and the thickness (d) (such as λ=α·d+β), the thickness is calculated directly through the peak offset. If the intensity ratio (such as I1 / I2) or full spectrum mode is used, a regression model (such as PLS partial least squares or neural network) is required to output the thickness pointer for spatial regionalization processing: first, regional division is performed; the processed surface is divided into N×N grids (such as 10×10), and each grid is independently matched with the spectrum and generates a thickness pointer; then, Kriging interpolation (Kriging) or adjacent area mean supplementation is used for low signal-to-noise ratio areas.

[0061] For each region’s thickness pointer sequence {d(t1), d(t2), ..., d(t n )} The instantaneous grinding rate is calculated by performing first-order difference (Δd / Δt); the exponentially weighted moving average (EWMA) or Kalman filter is used to eliminate short-term fluctuations and extract the stable rate curve.

[0062] A response model between grinding rate and processing parameters (pressure, speed, grinding fluid concentration) is established, and abnormality detection is enabled at the same time. When the rate threshold (such as ±10% deviation) is set, an alarm is triggered or the machine is automatically shut down.

[0063] Then, target thickness comparison and deviation control are performed. First, the preset thickness distribution is determined, and the objective function is adjusted according to the processing stage (rough grinding, fine grinding). Different weight coefficients are assigned to the edge area (prone to over-grinding) and the center area (prone to residue). The processing parameters are dynamically adjusted: If there are conflicts in multiple regions (such as insufficient center speed but excessive edge wear), the particle swarm algorithm (PSO) is used to solve the global optimal parameter combination.

[0064] The carrier head's zoned pressure control system dynamically adjusts grinding pressure in different areas, ensuring consistent grinding endpoint times across all areas to avoid over-grinding and residual material. The pointer trajectory is continuously monitored, and pressure parameters are updated in real time until the global endpoint condition is reached.

[0065] Example 2

[0066] The design of a magnesium oxide grinding monitoring system of the present invention is based on the method in Example 1, such as Figure 2 The process shown is as follows Figure 3 The following modules are shown: data acquisition module, data fusion module, automatic adjustment module;

[0067] The data acquisition module acquires multi-dimensional dynamic data of the grinding process in real time, providing raw input for subsequent analysis;

[0068] The data fusion module is used to comprehensively allocate the reflection spectrum weight coefficient obtained by each optical sensor according to the angle correction factor and the screen dust influence coefficient; and fuse the reflection spectra obtained by each optical sensor according to the reflection spectrum weight coefficient obtained by each optical sensor and output it;

[0069] The data fusion module includes a dust trajectory calculation module and an image dust impact calculation module;

[0070] The dust trajectory calculation module predicts the dust diffusion trajectory during the grinding process, determines the optical sensor capture range interval, and compares the optical sensor capture range interval with the predicted dust movement trajectory; if the optical sensor is blocked, K=1, otherwise, K=0;

[0071] The screen dust impact calculation module calculates the screen dust impact coefficient of each optical sensor according to the K value and the temporal and spatial concentration distribution of dust;

[0072] The automatic adjustment module dynamically controls the grinding parameters based on the fused data to ensure that the grinding endpoint time of each area is consistent; the real-time reflection spectrum is matched with the historical spectrum library, and the "thickness pointer" of each area is generated through principal component analysis; the target thickness is compared with the real-time thickness pointer to calculate the grinding rate deviation of each area; and differentiated grinding pressure is applied to different areas through the carrier head partition pressure control system.

[0073] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0074] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0075] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0076] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0077] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. A method for monitoring the grinding of magnesium oxide, characterized in that: The following steps are involved: collecting a reference spectrum generated by light reflection from the surface of a reference substrate having a target thickness; During the grinding process, multiple optical sensors are used to capture the surface reflection spectrum of the magnesium oxide grinding substrate at multiple angles; Calculate the dust impact coefficient of each optical sensor based on the K value and the temporal and spatial concentration distribution of dust; Determine the angle correction factor and the screen dust influence coefficient to comprehensively allocate the reflection spectrum weight coefficient obtained by each optical sensor; fuse the reflection spectra obtained by each optical sensor according to the reflection spectrum weight coefficient obtained by each optical sensor and output; The real-time spectrum is matched with the historical spectrum library to generate thickness indicators for each area, which are real-time thickness estimates. The target thickness of each area is compared to determine the rate deviation. The carrier head zone pressure control system dynamically adjusts the grinding pressure in different areas to ensure that the grinding endpoint time of all areas is close to the same. The K value is used to determine the capture range of the optical sensor, and the capture range of the optical sensor is compared with the predicted dust movement trajectory; If the optical sensor is blocked, K=1, otherwise, K=0.

2. The method for monitoring the grinding of magnesium oxide according to claim 1, wherein: The dust movement trajectory prediction is achieved through the Euler-Lagrangian coupling model, including the following steps: first, a three-dimensional geometric calculation domain is constructed based on the CAD model of the grinding equipment, and the airflow field is characterized by using the Navier-Stokes equation combined with the turbulence model; The force and motion trajectory of dust particles are calculated using the drag model and lift model.

3. The method for monitoring the grinding of magnesium oxide according to claim 1, wherein: The dust impact coefficient of each optical sensor is calculated using the logistic regression formula based on the K value and the spatiotemporal concentration of dust.

4. The method for monitoring the grinding of magnesium oxide according to claim 3, wherein: The dust spatiotemporal concentration refers to the spatial dust concentration at different times.

5. The method for monitoring the grinding of magnesium oxide according to claim 3, wherein: The reflection spectrum weight coefficient obtained by each optical sensor is comprehensively calculated based on the angle correction factor and the image dust influence coefficient.

6. The method for monitoring the grinding of magnesium oxide according to claim 5, wherein: The angle correction factor is determined by the following steps: first, a theoretical light intensity-polarization relationship is established based on the Fresnel reflection model; then, the actual angle of the sensor layout is calibrated by laser, and dynamic interference is compensated by combining the dust scattering coefficient and the optical path length; finally, the ratio of the measured reflectivity to the theoretical reflectivity is used as the angle correction factor.

7. The method for monitoring the grinding of magnesium oxide according to claim 5, wherein: The reflection spectra obtained by each optical sensor are fused and output according to the reflection spectrum weight coefficients obtained by each optical sensor.

8. The method for monitoring the grinding of magnesium oxide according to claim 1, wherein: The historical spectral library is constructed by the following method: first, collecting standard reflectance spectra of magnesium oxide at different thicknesses; then extracting characteristic peak positions, intensity ratios and full spectrum patterns as spectral features; and finally generating a low-dimensional feature library through principal component analysis.

9. A magnesium oxide grinding monitoring system, characterized in that: The monitoring system is based on the method described in any one of claims 1 to 8, and comprises the following modules: a data acquisition module, a data fusion module, and an automatic adjustment module; The data acquisition module acquires multi-dimensional dynamic data of the grinding process in real time, providing raw input for subsequent analysis; The data fusion module is used to comprehensively allocate the reflection spectrum weight coefficient obtained by each optical sensor according to the angle correction factor and the screen dust influence coefficient; and fuse the reflection spectra obtained by each optical sensor according to the reflection spectrum weight coefficient obtained by each optical sensor and output it; The automatic adjustment module dynamically controls the grinding parameters based on the fused data to ensure that the grinding endpoint time of each area is consistent; the real-time reflection spectrum is matched with the historical spectrum library, and the real-time thickness estimation value of each area is generated through principal component analysis; the target thickness is compared with the real-time thickness pointer to calculate the grinding rate deviation of each area; and differentiated grinding pressure is applied to different areas through the carrier head partition pressure control system.

10. A magnesium oxide grinding monitoring system according to claim 9, characterized in that: The data fusion module includes a dust trajectory calculation module and an image dust impact calculation module; The dust trajectory calculation module predicts the dust diffusion trajectory during the grinding process, determines the optical sensor capture range interval, and compares the optical sensor capture range interval with the predicted dust movement trajectory; If the optical sensor is blocked, K=1, otherwise, K=0; The image dust impact calculation module calculates the image dust impact coefficient of each optical sensor according to the K value and the spatiotemporal concentration distribution of dust.

Images