A system and method for online detection of polishing liquid particles during diamond polishing

Abstract

The present invention relates to the technical field of material detection, and specifically to an online detection system and method for polishing liquid particles during a diamond polishing process. The system comprises the following steps: identifying particle size information of particles in the polishing liquid through a particle size recognition model; dynamically exciting particles in a detection area in combination with different optical excitation conditions; synchronously collecting a first excitation response, a second excitation response, and a third excitation response generated by the particles after the dynamic excitation; statically exciting the particles under a fourth excitation condition, and detecting a fourth excitation response generated by the particles after the static excitation; combining the first excitation response, the second excitation response, the third excitation response, and the fourth excitation response to form a multimodal particle feature vector; matching the multimodal particle feature vector with a preset diamond model to identify diamond particles and distinguish non-diamond impurity particles, and outputting concentration and particle size distribution information of the diamond particles.

Description

Technical Field

[0001] The present invention relates to the technical field of material detection, and in particular to an online detection system and method for polishing liquid particles in a diamond polishing process. Background Art

[0002] Currently, the detection of diamond particle status in polishing fluids primarily relies on offline detection methods, such as laser particle size analyzers and dynamic light scattering. While these methods can provide particle size information with a certain degree of accuracy, they suffer from poor real-time performance, susceptibility to sample sedimentation and environmental interference, and an inability to identify particle material. Reusable polishing fluids may contain non-diamond impurity particles, such as silicon carbide and aluminum oxide. These particles resemble diamond particles in refractive index, particle size, and morphology, making them difficult to distinguish using conventional detection techniques, thus compromising polishing fluid quality assessment and performance.

[0003] Furthermore, while diamond and other high-refractive-index particles exhibit microscopic differences in light scattering behavior, polarization response, and spectral absorption characteristics, traditional detection methods often fail to fully utilize these physical properties for high-precision identification. Consequently, there is an urgent need for a detection system that can online identify and quantitatively analyze different particle components in polishing fluids based on their optical, scattering, or other physical properties.

[0004] Therefore, an online detection system and method for polishing liquid particles in a diamond polishing process are proposed. Summary of the Invention

[0005] The present invention aims to provide an online detection system and method for polishing slurry particles during diamond polishing to distinguish diamond particles from non-diamond impurities in the polishing slurry, ensuring that only diamond abrasive particles are identified and measured. The system includes: a particle size recognition module that identifies the particle size information of particles in the polishing slurry using a particle size recognition model; a dynamic excitation module that dynamically excites particles in the detection area using different optical excitation conditions; a dynamic excitation response acquisition module that synchronously collects the first, second, and third excitation responses generated by the particles after dynamic excitation; a static excitation and detection module that statically excites the particles under a fourth excitation condition and detects the fourth excitation response generated by the particles after static excitation; and a feature analysis and identification module that combines the first, second, third, and fourth excitation responses to form a multimodal particle feature vector. The multimodal particle feature vector is matched with a preset diamond model to identify diamond particles and distinguish non-diamond impurity particles, outputting information on the concentration and particle size distribution of the diamond particles.

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

[0007] An online detection system for polishing liquid particles during diamond polishing, comprising:

[0008] a particle size recognition module, configured to guide the polishing liquid to the detection area at a preset rate and identify the particle size information of the particles in the polishing liquid through a particle size recognition model;

[0009] a dynamic excitation module for dynamically exciting the particles in the detection area by combining different optical excitation conditions; the optical excitation conditions include a first excitation condition, a second excitation condition, and a third excitation condition; the third excitation condition is dynamically adjusted based on the particle size information;

[0010] A dynamic excitation response acquisition module, configured to synchronously acquire a first excitation response, a second excitation response, and a third excitation response generated by particles after the dynamic excitation;

[0011] a static excitation and detection module, configured to set a fourth excitation condition to statically excite the particles, and detect a fourth excitation response generated by the particles after the static excitation;

[0012] The feature analysis and identification module is used to combine the first excitation response, the second excitation response, the third excitation response and the fourth excitation response to form a multimodal particle feature vector; match the multimodal particle feature vector with a preset diamond model, identify diamond particles and distinguish non-diamond impurity particles, and output the concentration and particle size distribution information of the diamond particles.

[0013] Preferably, the particle size recognition model includes: a particle information acquisition unit, a motion trajectory extraction unit, an edge detection and contour reconstruction unit, and a particle size information estimation and calibration unit;

[0014] The particle information collection unit continuously collects particle information of particles moving with the liquid flow in the polishing liquid to obtain a particle information sequence;

[0015] The motion trajectory extraction unit calculates the particle velocity and relative residence time based on the movement trajectory of the particle in the particle information sequence;

[0016] The edge detection and contour reconstruction unit extracts the edge of the particle based on the particle velocity and relative residence time, and fits the two-dimensional projection contour of the particle;

[0017] The particle size information estimation and calibration unit calculates the edge contour area of the particle according to the two-dimensional projection contour, and identifies the particle size information of the particle in combination with preset optical calibration parameters.

[0018] Preferably, the first excitation condition is to set M different laser wavelengths for irradiation; the second excitation condition is to set N different laser incident angles for irradiation; and the third excitation condition includes a first polarization angle, a second polarization angle and a third polarization angle.

[0019] Preferably, the third excitation condition is dynamically adjusted based on the particle size information, specifically including:

[0020] If the particle size information is less than a preset first threshold, selecting a first polarization angle;

[0021] If the particle size information is greater than and / or equal to a preset first threshold and less than and / or equal to a preset second threshold, selecting a second polarization angle;

[0022] If the particle size information is greater than a preset second threshold, a third polarization angle is selected.

[0023] Preferably, the fourth excitation condition is to use an ultraviolet light source with a preset wavelength for continuous irradiation to excite the fluorescence response of the particles to obtain a fourth excitation response.

[0024] Preferably, the feature analysis and identification module includes: a particle feature extraction unit, a multimodal particle feature fusion unit, a model matching and classification unit, and a statistical output unit;

[0025] The particle feature extraction unit is used to extract characteristic parameters representing the optical properties of the particles from the first excitation response, the second excitation response, the third excitation response and the fourth excitation response to obtain the particle features;

[0026] The multimodal particle feature fusion unit performs weighted combination on the extracted particle features to form the multimodal particle feature vector;

[0027] The model matching and classification unit obtains the preset diamond model based on training data of known diamond particles and typical impurity particles, matches the input multimodal particle feature vector with the diamond model, and outputs the classification results and confidence levels of the particles as diamonds and non-diamond impurities;

[0028] The statistical output unit collects statistics and outputs the concentration and particle size distribution information of the diamond particles within the detection time period in real time according to the classification result and the confidence level.

[0029] Preferably, a method for online detection of polishing liquid particles during diamond polishing comprises:

[0030] The polishing liquid is directed to the detection area at a preset rate, and the particle size information of the particles in the polishing liquid is identified by a particle size recognition model;

[0031] Dynamically exciting the particles in the detection area in combination with different optical excitation conditions; the optical excitation conditions include a first excitation condition, a second excitation condition, and a third excitation condition; the third excitation condition is dynamically adjusted based on the particle size information;

[0032] Synchronously collecting a first excitation response, a second excitation response, and a third excitation response generated by the particles after the dynamic excitation;

[0033] Setting a fourth excitation condition to statically excite the particles, and detecting a fourth excitation response generated by the particles after the static excitation;

[0034] The first excitation response, the second excitation response, the third excitation response and the fourth excitation response are combined to form a multimodal particle feature vector; the multimodal particle feature vector is matched with a preset diamond model to identify diamond particles and distinguish non-diamond impurity particles, and the concentration and particle size distribution information of the diamond particles is output.

[0035] Compared with the prior art, the present invention has the following beneficial effects:

[0036] 1. This invention proposes a particle size recognition model that combines a multi-unit collaborative recognition mechanism, including trajectory analysis, edge detection, and particle size estimation, to achieve online, high-precision identification of particle size in polishing fluids, providing advance information support for the subsequent personalized setting of excitation conditions. Compared with traditional methods based on single scattering images or fluid dynamics-based particle size inference, this invention significantly reduces recognition error and has strong adaptability. It can maintain high resolution and recognition accuracy, especially in dynamic environments with mixed particles and strong background interference.

[0037] 2. The present invention introduces a multi-channel excitation mechanism and combines a dynamic excitation module with a static excitation module to collect scattering responses under different wavelengths, incident angles, polarization conditions and ultraviolet excitation conditions, significantly enhancing the optical feature dimension of particle identification. It can dynamically adjust the excitation parameters to adapt to the response characteristics of particles of different particle sizes. In the presence of optical interference or high-refractive-index impurities, it can still efficiently identify diamond particles, effectively solving the problem in existing solutions where diamond, aluminum oxide, and silicon carbide particles are difficult to distinguish due to their similar optical properties.

[0038] 3. This invention constructs multimodal particle feature vectors and introduces a pre-defined diamond recognition model based on machine learning training. This model can fuse and analyze the high-dimensional features of various excitation responses, significantly improving the classification accuracy of diamond and non-diamond impurities and the system's real-time response capabilities. Compared with existing solutions that rely on a single response parameter for pattern matching, this invention achieves millisecond-level response output while maintaining high accuracy, making it particularly suitable for online, continuous monitoring of polishing fluid status in industrial scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 A schematic structural diagram of an online detection system for polishing liquid particles during diamond polishing provided by an embodiment of the present invention;

[0040] Figure 2 A schematic flow chart of an online detection method for polishing liquid particles during diamond polishing provided by an embodiment of the present invention;

[0041] Figure 3 A schematic diagram of the structure of a particle size recognition model provided by an embodiment of the present invention;

[0042] Figure 4 This is a diagram showing the working principle of the feature analysis and recognition module provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0043] 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.

[0044] Diamond polishing is a high-precision surface treatment technology widely used in fields such as optics, electronics, and precision machinery. Its core is to use diamond particles in the polishing fluid to grind the workpiece surface to achieve a high-quality surface finish. The particle size and distribution of diamond particles in the polishing fluid directly affect the polishing effect and are key factors in determining polishing quality. However, in actual operation, non-diamond impurity particles are often mixed into the polishing fluid. These impurities can interfere with the polishing process, leading to surface defects or quality degradation. Therefore, real-time and accurate detection of diamond particle concentration and particle size distribution in the polishing fluid, while effectively distinguishing non-diamond impurity particles, has become an important technical requirement for ensuring polishing quality.

[0045] The present invention provides a system and method for online detection of polishing slurry particles during diamond polishing. These systems effectively distinguish diamond particles from non-diamond impurities in the polishing slurry during online detection, ensuring that only diamond abrasive particles are identified and measured. To demonstrate the effectiveness of the present method in distinguishing diamond particles from non-diamond impurities in the polishing slurry, two examples are presented below to illustrate the effectiveness of the present method.

[0046] Example 1

[0047] In the embodiment of the present application, the method proposed by the present invention is used to describe in detail the process of effectively distinguishing diamond particles from non-diamond impurities in the polishing liquid. The embodiment of the present application is suitable for online and continuous detection of the polishing liquid state during the diamond polishing process in industrial scenarios. Figure 1 The content describes in detail the online and continuous detection process of the polishing fluid state during diamond polishing; Figure 1 This is a specific structural diagram of the system proposed in the present invention, including: a particle size recognition module, which identifies the particle size information of particles in the polishing liquid through a particle size recognition model; a dynamic excitation module, which dynamically excites particles in the detection area in combination with different optical excitation conditions; a dynamic excitation response acquisition module, which synchronously collects the first excitation response, second excitation response, and third excitation response generated by the particles after dynamic excitation; a static excitation and detection module, which sets a fourth excitation condition to statically excite the particles and detects the fourth excitation response generated by the particles after static excitation; a feature analysis and recognition module, which combines the first excitation response, second excitation response, third excitation response, and fourth excitation response to form a multimodal particle feature vector; matches the multimodal particle feature vector with a preset diamond model, identifies diamond particles and distinguishes non-diamond impurity particles, and outputs the concentration and particle size distribution information of the diamond particles. Figure 2 This is a specific flow chart of the method proposed in the present invention. Figure 1 and Figure 2 The following describes the contents:

[0048] a particle size recognition module, configured to guide the polishing liquid to the detection area at a preset rate and identify the particle size information of the particles in the polishing liquid through a particle size recognition model;

[0049] The particle size recognition model includes: a particle information acquisition unit, a motion trajectory extraction unit, an edge detection and contour reconstruction unit, and a particle size information estimation and calibration unit;

[0050] The particle information collection unit continuously collects particle information of particles moving with the liquid flow in the polishing liquid to obtain a particle information sequence;

[0051] The motion trajectory extraction unit calculates the particle velocity and relative residence time based on the movement trajectory of the particle in the particle information sequence;

[0052] The edge detection and contour reconstruction unit extracts the edge of the particle based on the particle velocity and relative residence time, and fits the two-dimensional projection contour of the particle;

[0053] The particle size information estimation and calibration unit calculates the edge contour area of the particle according to the two-dimensional projection contour, and identifies the particle size information of the particle in combination with preset optical calibration parameters.

[0054] Specifically, the polishing liquid containing the particles to be tested is first guided to a transparent detection area at a stable and preset flow rate through a precision pump control. In this detection area, a particle information acquisition unit, including a light source (such as an LED strobe light) and a high-speed imaging device (such as a CMOS camera with a frame rate of hundreds to thousands of FPS), continuously captures the particles flowing through, forming an image sequence containing particle motion and morphological information, and obtaining a particle information sequence.

[0055] The motion trajectory extraction unit processes the particle information sequence and uses a particle image velocimetry algorithm to identify the position of the same particle in different frames of the particle information sequence image, thereby accurately calculating the motion speed of each particle and its relative residence time in the camera field of view;

[0056] Based on the calculated particle velocity and relative residence time, a multi-frame fusion algorithm (in this embodiment, multiple frames of images are aligned based on the particle trajectory and weighted averaged) is applied to the edge detection and contour reconstruction unit to improve image clarity. Next, a clear edge of the particle is extracted using an edge detection algorithm (such as the Canny operator, the Sobel operator, or a machine learning-based edge detection model), and a two-dimensional projection contour of the particle is further obtained using contour fitting techniques (such as minimum circumscribed circle, ellipse fitting, or irregular contour reconstruction).

[0057] The particle size estimation and calibration unit calculates particle size information, including geometric parameters such as area, equivalent diameter, and major and minor axes, based on the reconstructed 2D particle projection profile. Combined with optical system calibration parameters pre-calibrated using standard-sized particles (e.g., NIST-traceable microspheres), the unit accurately estimates the size of each particle.

[0058] Table 1 is a comparison table of particle size identification errors based on different methods.

[0059] Table 1 Comparison of particle size identification errors

[0060]

[0061] The particle size recognition module in this embodiment uses dynamic tracking and multi-frame processing technology to accurately measure the particle size of flowing particles in the polishing fluid online and in real time. This effectively overcomes motion blur, even at a certain flow rate, ensuring accurate particle size measurement. The resulting precise particle size information is not only a key parameter for assessing the polishing fluid's condition (e.g., whether the abrasive size is acceptable and whether there are abnormally large particles), but also provides important input for subsequent modules (such as the adaptive adjustment of the polarization angle in the dynamic excitation module), thereby enhancing the overall intelligence and accuracy of the detection system.

[0062] Preferably, a dynamic excitation module is used to dynamically excite the particles in the detection area in combination with different optical excitation conditions; the optical excitation conditions include a first excitation condition, a second excitation condition and a third excitation condition; the third excitation condition is dynamically adjusted based on the particle size information;

[0063] The first excitation condition is to set M different laser wavelengths for irradiation; the second excitation condition is to set N different laser incident angles for irradiation; the third excitation condition includes a first polarization angle, a second polarization angle and a third polarization angle.

[0064] Specifically, the first excitation condition is to use a programmable light source or filter wheel to switch, and use M lasers of different wavelengths (for example, M can be set to 5, such as 405nm, 488nm, 532nm, 633nm, 785nm) to irradiate the particles one by one to obtain the scattering or absorption characteristics of the particles under different laser wavelengths; the second excitation condition is to adjust the reflector in the laser light path or use a spatial light modulator to set N different laser incident angles (for example, N can be set to 3, such as 30 degrees, 45 degrees, 60 degrees) to irradiate the particles to detect the anisotropic scattering characteristics of the particles; the third excitation condition uses a liquid crystal variable phase retarder or a rotating polarizer combination to generate polarized light with a first polarization angle (for example, 0 degree), a second polarization angle (for example, 45 degrees) and a third polarization angle (for example, 90 degrees) to irradiate the particles, and the specific polarization angle selection of this third excitation condition will be dynamically adjusted based on the particle size information pre-obtained by the particle size recognition module.

[0065] By combining excitation with a variety of wavelengths, incident angles, and dynamically adjusted polarization angles, the optical properties of particles can be detected from multiple dimensions. Particles of different materials and morphologies respond differently to these excitation conditions. This multi-parameter, dynamic excitation method significantly enhances the system's ability to obtain characteristic information about particles, providing a richer and more discernible data foundation for subsequent precise differentiation of diamond particles from non-diamond impurities, improving both sensitivity and accuracy.

[0066] Preferably, the third excitation condition is dynamically adjusted based on the particle size information, specifically including:

[0067] If the particle size information is less than a preset first threshold, selecting a first polarization angle;

[0068] If the particle size information is greater than and / or equal to a preset first threshold and less than and / or equal to a preset second threshold, selecting a second polarization angle;

[0069] If the particle size information is greater than a preset second threshold, a third polarization angle is selected.

[0070] Specifically, a first threshold (e.g., 1 μm) and a second threshold (e.g., 5 μm);

[0071] When the particle size recognition module detects that the particle size is less than a preset first threshold, the dynamic excitation module automatically selects the first polarization angle (0-degree linear polarization) for excitation. If the detected particle size is greater than or equal to the first threshold and less than or equal to the second threshold, the second polarization angle (45-degree linear polarization) is selected for excitation. If the particle size is greater than the second threshold, the third polarization angle (90-degree linear polarization) is selected for excitation.

[0072] This embodiment dynamically adjusts the polarization angle for particles of varying size, enabling more effective detection of polarization-related optical properties, such as depolarization. This adaptive excitation strategy ensures high-quality polarization scattering signals for particles of all sizes, thereby improving the effectiveness of feature extraction and further enhancing the accuracy of distinguishing diamond from impurity particles, particularly those with similar morphology or material but different sizes.

[0073] Preferably, the dynamic excitation response acquisition module is used to synchronously acquire the first excitation response, the second excitation response and the third excitation response generated by the particles after the dynamic excitation;

[0074] Specifically, the excitation response is a scattered light spot image or an angle-resolved scattering pattern generated by the particles under each specific excitation condition.

[0075] Preferably, the static excitation and detection module is used to set a fourth excitation condition to statically excite the particles, and detect a fourth excitation response generated by the particles after the static excitation;

[0076] The fourth excitation condition is to use an ultraviolet light source with a preset wavelength for continuous irradiation to excite the fluorescence response of the particles to obtain a fourth excitation response.

[0077] Specifically, a continuous ultraviolet light source, such as one with a wavelength of 365 nm, is used to excite the particles, prompting them to produce a stable fluorescence signal. An acquisition system captures the excitation image of the particles under UV illumination and detects the corresponding fluorescence signal as a fourth excitation response, which complements and verifies the particle optical characteristics acquired under dynamic excitation conditions.

[0078] Table 2 is a comparison table of diamond and impurity recognition rates under different excitation conditions.

[0079] Table 2 Comparison of diamond and impurity recognition rates under different excitation conditions

[0080]

[0081] The introduction of fluorescence detection based on the intrinsic spectral properties of the material as a fourth excitation response greatly enhances the system's identification capabilities. Diamond and many potential non-diamond impurities (such as certain polymer particles, organic contaminants, and other mineral particles) often exhibit significant differences in fluorescence properties (e.g., luminescence, emission wavelength, fluorescence efficiency, and fluorescence lifetime) under ultraviolet or specific visible light excitation. This complementary detection method provides powerful chemical or structural fingerprint information independent of scattering properties for distinguishing diamond from non-diamond impurities, further improving the accuracy and robustness of the system's identification, especially for complex particle mixtures that are difficult to distinguish based on scattering properties alone.

[0082] Preferably, the feature analysis and identification module includes: a particle feature extraction unit, a multimodal particle feature fusion unit, a model matching and classification unit, and a statistical output unit;

[0083] The particle feature extraction unit is used to extract characteristic parameters representing the optical properties of the particles from the first excitation response, the second excitation response, the third excitation response and the fourth excitation response to obtain the particle features;

[0084] The multimodal particle feature fusion unit performs weighted combination on the extracted particle features to form the multimodal particle feature vector;

[0085] The model matching and classification unit obtains the preset diamond model based on training data of known diamond particles and typical impurity particles, matches the input multimodal particle feature vector with the diamond model, and outputs the classification results and confidence levels of the particles as diamonds and non-diamond impurities;

[0086] The statistical output unit collects statistics and outputs the concentration and particle size distribution information of the diamond particles within the detection time period in real time according to the classification result and the confidence level.

[0087] Specifically, the particle feature extraction unit is responsible for deep processing of the three dynamic excitation responses obtained from the scattering image acquisition module and the fourth excitation response obtained from the static excitation and detection module. Quantitative parameters such as the total scattering intensity, the integral of the scattering intensity within a specific angle range, the statistical moments of the scattering angle distribution (mean, variance, skewness, kurtosis), image texture features (such as contrast, energy, entropy based on the gray-level co-occurrence matrix), polarization degree, depolarization ratio, and particle projection shape descriptors (such as roundness and elongation) are extracted from the first, second, and third excitation responses; characteristic parameters such as the average fluorescence intensity, fluorescence spectrum peak wavelength, and peak area are extracted from the fourth excitation response. These parameters together constitute the particle features that characterize the multi-dimensional optical properties of a single particle.

[0088] The multimodal particle feature fusion unit first normalizes or standardizes features of different dimensions and numerical ranges. It then uses feature selection algorithms (such as principal component analysis (PCA), mutual information, and model-based feature importance ranking) to eliminate redundant or low-contribution features. It then intelligently weights and combines each feature based on its ability to distinguish diamond from impurities (through machine learning and adaptive learning), ultimately forming a multimodal particle feature vector.

[0089] The model matching and classification unit first constructs a preset diamond model based on a multimodal feature database of a large number of labeled known diamond particles and various typical non-diamond impurity particles; then, through a supervised machine learning algorithm (such as deep neural network DNN, support vector machine SVM, random forest RF, gradient boosting decision tree GBDT, etc.), the multimodal particle feature vector is matched with the preset diamond model, and the classification result and confidence level of whether the particle is judged to be diamond or non-diamond impurity are output.

[0090] By systematically extracting and intelligently integrating multi-source heterogeneous optical features and applying advanced data-driven machine learning models for matching and classification, the system not only accurately identifies individual particle types (diamonds or impurities) in the polishing slurry, but also outputs highly reliable particle identification results in real time, effectively distinguishing diamond particles from impurity particles, meeting the high-performance requirements of industrial-grade polishing slurry status monitoring. Ultimately, the real-time statistical information on diamond concentration and particle size distribution provides key online quality control parameters for the production process, helping to promptly detect changes in polishing slurry performance, optimize polishing process parameters, and ensure high quality and consistency of the final product.

[0091] The present invention provides an online detection system for polishing slurry particles during diamond polishing. This system features multiple modules working in tandem, effectively distinguishing diamond particles from non-diamond impurities in the polishing slurry during online detection, ensuring that only diamond abrasive particles are identified and measured. A particle size recognition module acquires particle size information and dynamically adjusts excitation conditions based on this information to achieve the optimal excitation response for particles of varying sizes. A dynamic excitation module combines multiple laser wavelengths, incident angles, and polarization angles to perform multi-dimensional excitation of particles. A dynamic excitation response acquisition module simultaneously collects response data under multiple excitation conditions. A static excitation and detection module utilizes specific ultraviolet laser irradiation to stimulate fluorescence responses, further enriching the particles' optical signature information. Subsequently, a feature analysis and identification module performs deep feature extraction and fusion on the first to fourth excitation responses, constructing a multimodal particle feature vector. This vector is then matched against a pre-set diamond model to accurately classify and identify diamond particles and non-diamond impurities. The system ultimately outputs particle concentration and size distribution information, providing data support for process optimization and quality control. This system can effectively improve the intelligent level of particle identification in polishing liquid, reduce the impact of impurity interference on polishing quality, and is widely applicable to diamond polishing processes in high-end fields such as optics, electronics and precision manufacturing.

[0092] Example 2

[0093] In Example 1, the system proposed by the present invention successfully achieved online, continuous detection of the polishing liquid state during the diamond polishing process in an industrial setting by effectively distinguishing diamond particles from non-diamond impurities in the polishing liquid. To further verify the effectiveness of the present invention, this application also proposed a method for online detection of polishing liquid particles during diamond polishing, and conducted online, continuous detection of polishing liquid particles in another diamond polishing experiment.

[0094] The polishing liquid is directed to the detection area at a preset rate, and the particle size information of the particles in the polishing liquid is identified by a particle size recognition model;

[0095] The particle size recognition model includes: a particle information acquisition unit, a motion trajectory extraction unit, an edge detection and contour reconstruction unit, and a particle size information estimation and calibration unit;

[0096] The particle information collection unit continuously collects particle information of particles moving with the liquid flow in the polishing liquid to obtain a particle information sequence;

[0097] The motion trajectory extraction unit calculates the particle velocity and relative residence time based on the movement trajectory of the particle in the particle information sequence;

[0098] The edge detection and contour reconstruction unit extracts the edge of the particle based on the particle velocity and relative residence time, and fits the two-dimensional projection contour of the particle;

[0099] The particle size information estimation and calibration unit calculates the edge contour area of the particle according to the two-dimensional projection contour, and identifies the particle size information of the particle in combination with preset optical calibration parameters.

[0100] Preferably, the particles in the detection area are dynamically excited in combination with different optical excitation conditions; the optical excitation conditions include a first excitation condition, a second excitation condition and a third excitation condition; the third excitation condition is dynamically adjusted based on the particle size information;

[0101] The first excitation condition is to set M different laser wavelengths for irradiation; the second excitation condition is to set N different laser incident angles for irradiation; the third excitation condition includes a first polarization angle, a second polarization angle and a third polarization angle.

[0102] Preferably, the third excitation condition is dynamically adjusted based on the particle size information, specifically including:

[0103] If the particle size information is less than a preset first threshold, selecting a first polarization angle;

[0104] If the particle size information is greater than and / or equal to a preset first threshold and less than and / or equal to a preset second threshold, selecting a second polarization angle;

[0105] If the particle size information is greater than a preset second threshold, a third polarization angle is selected.

[0106] Preferably, the first excitation response, the second excitation response and the third excitation response generated by the particles after the dynamic excitation are collected synchronously.

[0107] Preferably, a fourth excitation condition is set to statically excite the particles, and a fourth excitation response generated by the particles after the static excitation is detected;

[0108] The fourth excitation condition is to use an ultraviolet light source with a preset wavelength for continuous irradiation to excite the fluorescence response of the particles to obtain a fourth excitation response.

[0109] Preferably, the first excitation response, the second excitation response, the third excitation response and the fourth excitation response are combined to form a multimodal particle feature vector; the multimodal particle feature vector is matched with a preset diamond model to identify diamond particles and distinguish non-diamond impurity particles, and the concentration and particle size distribution information of the diamond particles is output.

[0110] The feature analysis and identification module includes: a particle feature extraction unit, a multimodal particle feature fusion unit, a model matching and classification unit, and a statistical output unit;

[0111] The particle feature extraction unit is used to extract characteristic parameters representing the optical properties of the particles from the first excitation response, the second excitation response, the third excitation response and the fourth excitation response to obtain the particle features;

[0112] The multimodal particle feature fusion unit performs weighted combination on the extracted particle features to form the multimodal particle feature vector;

[0113] The model matching and classification unit obtains the preset diamond model based on training data of known diamond particles and typical impurity particles, matches the input multimodal particle feature vector with the diamond model, and outputs the classification results and confidence levels of the particles as diamonds and non-diamond impurities;

[0114] The statistical output unit collects statistics and outputs the concentration and particle size distribution information of the diamond particles within the detection time period in real time according to the classification result and the confidence level.

[0115] Table 3 is a comparison table of recognition accuracy of different methods.

[0116] Table 3 Recognition accuracy comparison table

[0117]

[0118] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. An online detection system for polishing liquid particles during diamond polishing, characterized in that: include: a particle size recognition module, configured to guide the polishing liquid to the detection area at a preset rate and identify the particle size information of the particles in the polishing liquid through a particle size recognition model; A dynamic excitation module is used to dynamically excite particles in the detection area by combining different optical excitation conditions; the optical excitation conditions include a first excitation condition, a second excitation condition, and a third excitation condition; the first excitation condition is to set M different laser wavelengths for irradiation; the second excitation condition is to set N different laser incident angles for irradiation; The third excitation condition includes a first polarization angle, a second polarization angle, and a third polarization angle; The third excitation condition is dynamically adjusted based on the particle size information, specifically including: if the particle size information is less than a preset first threshold, selecting a first polarization angle; if the particle size information is greater than and / or equal to the preset first threshold and less than and / or equal to a preset second threshold, selecting a second polarization angle; if the particle size information is greater than the preset second threshold, selecting a third polarization angle; A dynamic excitation response acquisition module is used to synchronously acquire the first excitation response, the second excitation response, and the third excitation response generated by the particles after dynamic excitation; A static excitation and detection module, configured to statically excite the particles under a fourth excitation condition and detect a fourth excitation response generated by the particles after the static excitation; the fourth excitation condition is to continuously irradiate the particles with an ultraviolet light source of a preset wavelength to excite the fluorescence response of the particles and obtain the fourth excitation response; The feature analysis and identification module includes a model matching and classification unit, which is used to combine the first excitation response, the second excitation response, the third excitation response and the fourth excitation response to form a multimodal particle feature vector; the model matching and classification unit matches the multimodal particle feature vector with a preset diamond model through a supervised machine learning algorithm, outputs the classification results and confidence levels of the particles as diamond and non-diamond impurities, identifies diamond particles and distinguishes non-diamond impurity particles; outputs the concentration and particle size distribution information of the diamond particles based on the classification results and confidence levels; the preset diamond model is trained based on known diamond particles and typical impurity particle data.

2. The system for online detection of polishing liquid particles during diamond polishing according to claim 1, characterized in that: The particle size recognition model includes: a particle information acquisition unit, a motion trajectory extraction unit, an edge detection and contour reconstruction unit, and a particle size information estimation and calibration unit; The particle information collection unit continuously collects particle information of particles moving with the liquid flow in the polishing liquid to obtain a particle information sequence; The motion trajectory extraction unit calculates the particle velocity and relative residence time based on the movement trajectory of the particle in the particle information sequence; The edge detection and contour reconstruction unit extracts the edge of the particle based on the particle velocity and relative residence time, and fits the two-dimensional projection contour of the particle; The particle size information estimation and calibration unit calculates the edge contour area of the particle according to the two-dimensional projection contour, and identifies the particle size information of the particle in combination with preset optical calibration parameters.

3. The online detection system for polishing liquid particles during diamond polishing according to claim 1, characterized in that: The feature analysis and identification module further includes: a particle feature extraction unit, a multimodal particle feature fusion unit and a statistical output unit; The particle feature extraction unit is used to extract characteristic parameters representing the optical properties of the particles from the first excitation response, the second excitation response, the third excitation response and the fourth excitation response to obtain the particle features; The multimodal particle feature fusion unit performs weighted combination on the extracted particle features to form the multimodal particle feature vector; The statistical output unit collects statistics and outputs the concentration and particle size distribution information of the diamond particles within the detection time period in real time according to the classification result and the confidence level.

4. A method for online detection of polishing liquid particles during diamond polishing, comprising executing the system for online detection of polishing liquid particles during diamond polishing as claimed in claim 1, characterized in that: include: The polishing liquid is directed to the detection area at a preset rate, and the particle size information of the particles in the polishing liquid is identified by a particle size recognition model; Dynamically exciting the particles in the detection area in combination with different optical excitation conditions; the optical excitation conditions include a first excitation condition, a second excitation condition, and a third excitation condition; the third excitation condition is dynamically adjusted based on the particle size information; Synchronously collecting a first excitation response, a second excitation response, and a third excitation response generated by the particles after the dynamic excitation; Setting a fourth excitation condition to statically excite the particles, and detecting a fourth excitation response generated by the particles after the static excitation; The first excitation response, the second excitation response, the third excitation response and the fourth excitation response are combined to form a multimodal particle feature vector; the multimodal particle feature vector is matched with a preset diamond model to identify diamond particles and distinguish non-diamond impurity particles, and the concentration and particle size distribution information of the diamond particles is output.

Images