An online detection system for diamond polishing and grinding

Abstract

The present invention relates to the field of solid analysis, and specifically to an online detection system for diamond polishing and grinding, comprising: first, synchronously acquiring optical sensor data and thermal sensor data of the diamond surface to be tested; then, processing the collected sensor data to detect and identify various physical properties of its surface, including surface defect status, thermally induced anomaly distribution, and self-calibrated surface roughness, and tracking the evolution trend of related properties, thereby generating preliminary multi-dimensional characteristic analysis results; finally, combining various analysis results to generate a comprehensive online evaluation report on the physical properties of the diamond defect surface. The present invention mainly solves the problem that traditional detection methods are difficult to fully evaluate the multi-dimensional physical properties of the surface in real time and accurately, and to quickly respond to changes in the surface state. It aims to comprehensively evaluate the surface quality of the diamond through automated online analysis, providing a basis for ensuring the quality of the final product and optimizing the grinding process.

Description

Technical Field

[0001] The invention relates to the technical field of solid analysis, in particular to an online detection system for diamond polishing and grinding. Background Art

[0002] Existing technologies for the precise measurement and in-depth analysis of diamond physical properties involve a variety of methods, aiming to quantitatively characterize key physical parameters such as surface topography, the size and distribution of microscopic defects, and optical properties, thereby evaluating the effectiveness of inspections. For example, specific optical configurations such as bright-field and dark-field illumination and differential interference contrast can enhance the detection and characterization of tiny scratches, pits, or contaminants, enabling physical parameter-based quality assessments.

[0003] However, applying these analytical techniques to online diamond inspection systems still faces numerous challenges. While traditional laboratory precision analytical instruments can provide high-resolution information on surface physical properties, they are slow and environmentally demanding, making them difficult to integrate directly into high-speed production lines exposed to vibration, coolants, and grinding dust for real-time physical property measurements. For high-hardness, highly reflective, or transparent materials like diamond, certain optical inspection methods may encounter low signal contrast, severe stray light interference, or limited penetration depth when measuring their surface physical properties, directly impacting the accuracy and reliability of the physical property measurements. Furthermore, online inspection systems must process large amounts of sensor data in a very short period of time and accurately identify and quantify a variety of tiny and diverse surface defects or parameter deviations. These technologies lack real-time performance, robustness, and intelligence. Therefore, most current diamond quality control still relies on offline spot checks, which not only results in delayed feedback on dynamic changes in the material's physical properties, making it difficult to achieve real-time optimization of the processing technology, but also increases the risk of scrap.

[0004] Therefore, an online detection system for diamond polishing and grinding is proposed. Summary of the Invention

[0005] The present invention aims to provide an online detection system for diamond polishing and grinding, which includes a defect data acquisition unit for synchronously acquiring optical sensing data and thermal sensing data of the diamond surface, and performing preprocessing operations including image enhancement, filtering, and digital feature extraction to output optical characterization data; a defect characteristic analysis unit, based on the optical characterization data, interprets image content through a fused data-driven and visual detection model to identify surface defect information, and performs image feature matching and time-series tracking to output corresponding defect evolution trend information; and a defect information synthesis unit, which utilizes a rule engine to generate a report on the physical characteristics of the defect surface to achieve closed-loop feedback, thereby realizing online, automated diamond surface condition evaluation and process optimization.

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

[0007] A diamond polishing and grinding online detection system, comprising:

[0008] Defect data acquisition unit, used to synchronously acquire optical sensing data and thermal sensing data of the diamond surface, and pre-process and output optical characterization data containing surface physical property information;

[0009] a defect characteristic analysis unit configured to detect defects on the diamond surface based on the optical characterization data by fusing a data-driven model with a visual detection model, output surface defect information, and detect thermal anomalies in combination with the thermal sensing data, and output thermal anomaly information; calculate the surface roughness of the diamond surface based on the optical characterization data, and output surface roughness information by comparing the surface roughness with external reference data and self-calibrating the surface roughness calculation; and compare the surface defect information with a historical defect database to output defect evolution trend information;

[0010] The defect information synthesis unit is used to make feedback decisions based on the surface defect information, the thermal anomaly information, the surface roughness information and the defect evolution trend information using a rule engine, generate a defect surface physical property report, and output it to the controller through an industrial communication interface.

[0011] Furthermore, the preprocessing specifically includes:

[0012] Converting the optical sensing data into a single-channel grayscale image;

[0013] applying a contrast-limited adaptive histogram equalization algorithm to the single-channel grayscale image;

[0014] Using a guided filtering algorithm to perform noise filtering on the contrast-enhanced single-channel grayscale image;

[0015] And for the filtered single-channel grayscale image, the region of interest is extracted by template matching technology based on normalized cross-correlation, and multi-level wavelet transform is applied to extract and reconstruct low-frequency approximate components and high-frequency detail components to form the optical characterization data.

[0016] Furthermore, the process of outputting surface defect information includes:

[0017] The data-driven model uses a lightweight convolutional neural network; using a lightweight convolutional neural network model that has been pre-trained for diamond surface defect types, it automatically extracts and identifies surface irregular defects from the input optical characterization data and outputs a first defect detection result, including a defect category, a prediction score with a confidence level greater than a preset threshold, and bounding box coordinates;

[0018] The visual inspection model integrates the Canny edge detection algorithm and the Hough transform algorithm to identify and locate regular surface defects and obtain a second defect detection result;

[0019] The first defect detection result and the second defect detection result are fused based on a confidence weighting factor and / or an intersection-over-union ratio between bounding boxes, and the surface defect information is generated through redundancy elimination and conflict resolution.

[0020] Furthermore, the process of outputting the surface roughness information includes:

[0021] Based on the optical characterization data, applying fast Fourier transform processing to analyze the spectral energy distribution of low-frequency approximate components and high-frequency detail components; and estimating the surface roughness and quantifying texture feature parameters based on a predetermined estimation model between the spectral energy distribution and the surface roughness, to obtain preliminary surface roughness information;

[0022] Periodically receiving a baseline surface roughness reference value for the same sample area via a calibration data interface; and applying an adaptive calibration algorithm to calculate a deviation by comparing the surface roughness with the baseline surface roughness reference value;

[0023] Based on the deviation, internal parameters of the predetermined estimation model are adjusted, the surface roughness is recalculated, and the calibrated surface roughness information is output.

[0024] Furthermore, the process of outputting the defect evolution trend information includes:

[0025] Correlation processing is performed on the surface defect information and the defect map data stored in the historical defect database in terms of time and space dimensions and matching of multi-dimensional characteristic parameters is performed to obtain an associated defect number;

[0026] Tracking the time-series quantitative change data corresponding to the defect number in the optical sensing data, and analyzing the time-series quantitative change data using a preset time series analysis model to identify and quantify the evolution pattern corresponding to the defect number; wherein the evolution pattern is a parameter vector that characterizes the dynamic statistical characteristics of the defect size change rate, morphological transformation characteristics, and occurrence frequency;

[0027] Based on the evolution pattern and a preset risk assessment rule set, the future development trend and potential risk level of each corresponding defect number are determined to obtain the defect evolution trend information.

[0028] Furthermore, the specific implementation process of the rule engine includes:

[0029] Analyzing the severity of the surface defect information and the degree of deviation of the surface roughness information from a preset target specification, and setting a processing priority queue;

[0030] The rule engine integrates the surface defect information, the defect evolution trend information, the surface roughness information, and the thermal anomaly information based on the processing priority queue, applies an adjustment action rule library, and generates a report on the physical properties of the defect surface.

[0031] Furthermore, the defect information synthesis unit also includes: using the isolation forest algorithm and the principal component analysis reconstruction error analysis method to compare the surface defect information, the thermal anomaly information and the surface roughness information with the normal operating condition model to identify new surface anomaly states, and the new surface anomaly states are used to trigger and update the rule engine to generate the defect surface physical property report.

[0032] Compared with the prior art, the present invention has the following beneficial effects:

[0033] 1. The present invention achieves synchronous, high-fidelity acquisition of optical and thermal sensing data from the diamond surface through a defect data acquisition unit. Combined with a sophisticated image preprocessing process including image enhancement, filtering, and targeted feature extraction, this method effectively overcomes the inherent problems of insufficient information dimensions and low signal-to-noise ratio inherent in traditional single information sources when comprehensively and precisely characterizing the physical properties of complex surfaces. The highly standardized optical characterization data output by this unit, which contains rich physical property information, also lays a high-quality, multimodal fusion data foundation for subsequent multi-dimensional intelligent analysis and precise defect interpretation.

[0034] 2. The defect characteristic analysis unit of this invention not only achieves efficient identification and accurate information output of various complex and regular defects on diamond surfaces by integrating data-driven image analysis models with machine vision algorithms, but also integrates analysis of thermal sensor data to identify potential thermally induced anomalies. It can accurately estimate surface roughness based on optical image features and achieve self-calibration by comparing with external benchmark data, ensuring the long-term accuracy of measurements. Furthermore, by intelligently comparing current defect information with a historical defect database and tracking its time series, it provides precise quantitative indicators for dynamically understanding the real-time quality status of the diamond surface.

[0035] 3. The defect information synthesis unit of this invention, through its built-in rule engine, enables rapid and accurate mapping from complex image analysis results to reports on the physical properties of specific defect surfaces, enabling intelligent decision-making. These adjustment suggestions are output to the equipment controller in real time via a standardized industrial communication interface, forming an agile closed-loop feedback loop from online detection to process execution. This not only improves the timeliness and pertinence of process adjustments, effectively suppressing quality fluctuations, but also drives the diamond polishing process towards higher automation, higher precision, and better resource utilization, thereby improving the final product qualification rate and the overall efficiency of the production line. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 A schematic structural diagram of a diamond polishing and grinding online detection system provided by the present invention;

[0037] Figure 2 A schematic flow chart of the surface defect information output process provided by the present invention;

[0038] Figure 3 This is a schematic flow chart of the surface roughness information output process provided by the present invention. DETAILED DESCRIPTION

[0039] 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.

[0040] See also Figures 1 to 3 The present invention provides an online detection system for diamond polishing and grinding, and the technical solution is as follows:

[0041] Example 1:

[0042] In order to effectively address the challenges of precise and real-time online monitoring of workpiece surface quality during diamond polishing and grinding, especially the difficulties of diverse defect morphologies, small feature sizes, and complex texture background interference, this embodiment proposes and elaborates a diamond polishing and grinding online detection system, such as Figure 1 As shown, including:

[0043] refer to Figure 1 The defect data acquisition unit is used to synchronously acquire optical sensing data and thermal sensing data of the diamond surface, and pre-process and output optical characterization data containing surface physical property information.

[0044] Specifically, in this embodiment, a high-speed industrial camera is integrated to capture high-resolution optical sensor data of the diamond workpiece. Simultaneously, the system is equipped with an infrared thermal imaging device that works synchronously with the optical camera to collect thermal sensor data from the same or adjacent areas, directly reflecting the real-time temperature distribution on the workpiece surface.

[0045] Furthermore, for the optical sensing data obtained from the industrial camera, the preprocessing specifically includes:

[0046] First, the optical sensor data captured by industrial cameras (typically in RGB format) is converted into a single-channel grayscale image. This process removes color information from the image, allowing subsequent analysis to focus on brightness and contrast, two physical properties closely related to surface defects (such as scratches, pits, and cracks) and texture features. Grayscaling effectively reduces data dimensionality and computational complexity, improving the real-time performance of subsequent processing.

[0047] Next, the contrast-limited adaptive histogram equalization (CLAHE) method is applied to the converted grayscale image. The CLAHE algorithm divides the image into multiple small grid regions (tiles) and performs local histogram equalization on each region, while also setting clipping limits to prevent excessive local contrast amplification. This method consistently enhances the contrast of image details. In images of diamond-polished surfaces, this method enhances the visibility of tiny scratches and pits in low-contrast areas, facilitating subsequent feature extraction and defect identification.

[0048] Subsequently, a guided filtering algorithm is used to remove noise from the contrast-enhanced single-channel grayscale image after CLAHE processing. This algorithm uses the image's structural features as a guide, effectively suppressing high-frequency interference such as coolant reflections, ambient light fluctuations, and sensor-generated image noise while preserving edge detail. This algorithm also preserves the continuity of key features such as microcrack edges and scratch outlines while suppressing noise.

[0049] Next, to focus analysis resources and improve processing efficiency, a template matching method based on normalized cross correlation (NCC) is used to dynamically extract regions of interest (ROIs) from the denoised image. This is matched block by block with a preset defect template image in the system, calculating a similarity score. The matching confidence level is then used to determine whether to extract a specific region for subsequent processing. If the matching result does not reach the set threshold, the entire image is used for processing by default to ensure detection integrity and robustness.

[0050] Finally, a multi-level discrete wavelet transform is applied to the extracted ROI or the entire image for feature extraction. The wavelet transform decomposes the image into low-frequency approximate components and high-frequency detail components in multiple directions. The former represents image contours and slowly varying regions, while the latter accurately captures information such as edges and texture mutations that represent defects in the image. In this embodiment, the wavelet basis can be selected from common forms such as "db4" or "sym4" based on the diamond surface texture characteristics. A 2- to 4-level decomposition is performed to construct a feature matrix of uniform size (e.g., 256×256), which facilitates subsequent recognition model invocation.

[0051] After preprocessing the diamond surface characteristics, the system can extract standardized feature image information with complete structure, retained details, concentrated features and moderate data volume from the original optical image, providing a high-quality data foundation for subsequent defect identification, roughness analysis and thermal anomaly detection, thereby improving the system's ability to identify tiny defects and the accuracy and stability of overall detection.

[0052] refer to Figure 1 The defect characteristic analysis unit is used to detect defects on the diamond surface based on the optical characterization data by fusing a data-driven model with a visual detection model, output surface defect information, and detect thermal anomalies in combination with the thermal sensing data, and output thermal anomaly information; calculate the surface roughness of the diamond surface based on the optical characterization data, and self-calibrate the surface roughness calculation by comparing it with external reference data, and output surface roughness information; and compare the surface defect information with a historical defect database to output defect evolution trend information.

[0053] Among them, this unit combines the input thermal sensing data or its derived characteristics to analyze whether there are thermal anomalies caused by factors such as processing friction, internal material stress or coating problems, such as local temperature exceeding the normal operating threshold (such as >50°C) and temperature gradient abnormality (such as >10°C / mm), and outputs the corresponding thermal anomaly information.

[0054] Furthermore, Figure 2 The figure is a flow chart of the surface defect information output process provided by the present invention. Figure 2 As shown in Figure 2, the process of outputting surface defect information includes:

[0055] The data-driven model uses a lightweight convolutional neural network (CNN), such as MobileNetV3 or EfficientNet-Lite. This model is pre-trained using supervised learning on a large amount of diamond surface image data, annotated with defect categories and bounding box locations, such as microcracks, pits, pits, contamination spots, and burn marks. Data augmentation techniques such as rotation, mirroring, and scaling are introduced during training to improve the model's robustness under varying lighting, viewing angles, and textures.

[0056] During the online detection phase, a lightweight convolutional neural network model pre-trained for diamond surface defect types is used to automatically extract and identify surface irregular defects from the input optical characterization data, and output a structured first defect detection result, including: defect category label, confidence score greater than a preset threshold, and defect bounding box coordinate information.

[0057] The visual inspection model integrates the Canny edge detection algorithm and the Hough transform method to identify and locate regular surface defects, such as linear scratches caused by abrasive particles or circular arc defects caused by chipped edges. The Canny algorithm extracts the edge structure of the image, while the Hough transform identifies geometric elements that conform to the parametric model from the edge map. The output includes the start and end coordinates, length, and angle information of the scratch, or the center coordinates, radius, start and end angles of the arc, thereby forming a secondary defect detection result.

[0058] To improve the completeness and accuracy of the recognition results, the first defect detection result and the second defect detection result are fused based on the confidence weighting factor and / or the intersection-over-union ratio between bounding boxes, and the surface defect information is generated through redundancy elimination and conflict resolution, specifically including:

[0059] Weighted fusion: The results are fused based on the confidence weight factors and / or bounding box intersection over union (IoU) of each detection source. For example, irregular defects identified by the CNN model can be given a higher initial weight; while regular defects have their fusion weight adjusted based on traditional features such as geometric structure integrity and edge strength.

[0060] Redundancy elimination: Detection boxes with high IoU overlap and low confidence between multiple detection results are eliminated to avoid repeated counting; for repeatedly identified defect information of the same type, they are merged and the one with higher confidence is retained.

[0061] Conflict resolution: When the same area is identified as different types of defects by two models, the system determines the most reasonable defect classification and description based on preset priority rules, typical morphological feature matching results, or introduces contextual features to assist in judgment.

[0062] The surface defect information should include each defect's unique identification ID, type classification, geometric location coordinates, size parameters (such as length, area or depth estimation), and optional morphological description features.

[0063] By integrating data-driven models with machine vision inspection algorithms, combining their respective strengths in irregular and regular defect detection, the comprehensiveness, accuracy, and stability of diamond surface defect detection have been enhanced. Leveraging an intelligent fusion processing mechanism, this approach not only achieves complementary advantages, overcoming the limitations of single detection methods in defect type coverage and recognition accuracy, but also ensures more reliable and complete surface defect information output through redundancy elimination and conflict resolution. This provides higher-quality input for subsequent quality assessment and process feedback, thereby more effectively guiding production and reducing the risk of misjudgment.

[0064] Further, Figure 3 The figure is a flow chart of the surface roughness information output process provided by the present invention. Figure 3 As shown, the process of outputting the surface roughness information includes:

[0065] First, based on the optical characterization data, fast Fourier transform (FFT) processing is applied to analyze the spectral energy distribution of the low-frequency approximate component and the high-frequency detail component. The spectral energy distribution includes but is not limited to statistical indicators such as frequency band energy proportion, main frequency position, and spectrum width.

[0066] Then, according to the spectrum energy distribution and the surface roughness (arithmetic mean roughness ) is used to estimate the surface roughness and quantify the texture feature parameters to obtain preliminary surface roughness information, wherein the texture feature parameters include but are not limited to indicators such as texture directionality, periodicity, contrast, and uniformity, and the surface roughness and quantified texture feature parameters together constitute the preliminary surface roughness information.

[0067] In addition, the predetermined estimation model is modeled based on the relationship between the image frequency domain features and the external physical measurement results to improve the accuracy and generalization ability of roughness estimation. The following can be selected:

[0068] Physical analytical model based on optical scattering theory;

[0069] Empirical models based on regression training of large sample data, such as linear regression models, support vector regression models, or lightweight neural network models.

[0070] Through a predefined calibration data interface, the system regularly receives (either periodically or on demand) a baseline surface roughness reference value for the same sample area. This reference value is typically measured by instruments such as an atomic force microscope (AFM), white light interferometer (WLI), or contact profilometer. It can be imported into the system via a manual input interface, file import, or network connection. Common calibration triggers include daily fixed cycles, batch changes, mold changes, or adjustments to key process parameters. The system applies an adaptive calibration algorithm to compare the surface roughness with the baseline surface roughness reference value, calculating the absolute or relative error and determining the deviation, which serves as the basis for model calibration.

[0071] Based on the deviation, internal parameters of the predetermined estimation model are adjusted, and the surface roughness is recalculated to output the calibrated surface roughness information, for example:

[0072] For linear models , the coefficients a, b and c can be modified according to the error feedback, where is the arithmetic mean roughness, and It is to extract the spectrum features that can reflect the roughness related information from the spectrum energy distribution;

[0073] For the lookup table model, the output value of the corresponding spectrum entry can be adjusted;

[0074] For neural networks or other nonlinear models, their weight parameters or bias terms can be updated.

[0075] On the one hand, the use of FFT to analyze the spectrum energy of image features can quickly and non-contactly obtain rich information related to the surface micromorphology, which provides a basis for online estimation. On the other hand, the self-calibration function, which regularly compares with high-precision physical measurement benchmark values ​​and automatically adjusts the estimated model parameters, effectively overcomes the defects of roughness measurement methods based solely on image analysis, which are easily affected by factors such as lighting changes, surface contamination, and insufficient model generalization ability, resulting in drift or systematic errors, and provides reliable feedback data for refined diamond process control.

[0076] Furthermore, the process of outputting the defect evolution trend information includes:

[0077] First, the surface defect information is associated with the defect map data stored in the historical defect database, and the associated defect numbers are obtained by performing temporal and spatial correlation processing and matching of multi-dimensional feature parameters.

[0078] During the association process, the system comprehensively considers the spatial proximity of the defect's spatial position, potential motion trajectory prediction, and corresponding processing stage timestamps, and at the same time matches multi-dimensional features such as the defect type, size, and shape factor to identify whether it belongs to the evolution of the same defect in the historical database, and then establishes a tracking relationship and obtains the associated defect number.

[0079] Among them, the historical defect database records the defect states observed for the same workpiece (such as repeated inspection at a fixed workstation) or the same batch of workpieces at multiple different inspection time points or processing stages, providing a reference basis for temporal evolution analysis.

[0080] Track the time-series quantitative change data corresponding to the defect number in the optical sensor data, including data on changes in geometric parameters (such as length, width, area, and depth), morphological parameters (such as aspect ratio, circularity, and edge complexity), and relative position parameters (such as distance from the edge or offset within the region) in the continuous inspection images over time or during the processing cycle. Analyze this time-series quantitative change data using a preset time series analysis model to identify and quantify the evolution pattern corresponding to the defect number. This model can be a traditional statistical method (such as ARIMA, adaptive exponential smoothing, or a Kalman filter) or a deep learning-based sequence model (such as an RNN or LSTM) to identify the evolution patterns of defects and extract their dynamic change characteristics.

[0081] Among them, the evolution pattern is a parameter vector that characterizes the defect size change rate (such as linear growth, exponential growth or stable state), morphological transformation characteristics such as point expansion to line, edge tending to irregularity) and dynamic statistical characteristics of occurrence frequency (such as periodic occurrence of defects or an increasing trend in occurrence frequency). This vector constitutes a mathematical representation of the defect's temporal evolution behavior.

[0082] Based on the evolution pattern and a preset risk assessment rule set, the future development trend and potential risk level of each corresponding defect number are determined to obtain the defect evolution trend information.

[0083] Among them, the risk assessment rule set can be established based on expert experience or historical big data regression analysis. For example: if the crack length growth rate exceeds the set threshold, the predicted future development trend is "rapid expansion" and the corresponding potential risk level is set to "high".

[0084] The system ultimately generates defect evolution trend information, including predictive future development trends and potential risk levels, which can be used to formulate subsequent diamond process adjustment strategies or maintenance plans. By intelligently correlating current inspection information with historical data and implementing time-series modeling, the system accurately quantifies the defect's rate of change, morphological evolution trends, and cyclical behavior. This enables forward-looking risk assessment, enabling preventive measures to be taken before defects evolve into unacceptable states. This reduces batch defect risk, improves diamond polishing reliability, and provides a data foundation for process optimization.

[0085] The defect information synthesis unit is used to make feedback decisions based on the surface defect information, the thermal anomaly information, the surface roughness information and the defect evolution trend information using a rule engine, generate a defect surface physical property report, and output it to the controller through an industrial communication interface.

[0086] Furthermore, the specific implementation process of the rule engine includes:

[0087] The rule engine first receives various output information from the defect characteristic analysis unit, mainly including:

[0088] The surface defect information includes defect type, location, size, quantity, severity assessment, etc.

[0089] The defect evolution trend information includes predictions of future development and risks of each defect;

[0090] The surface roughness information after calibration: such as Ra value and its deviation from the target specification;

[0091] And the thermal anomaly information: such as the maximum temperature and area range of the local overheating area.

[0092] Before processing this input information, the system first evaluates the severity of the surface defect (such as whether it is a critical defect, whether the size exceeds the threshold, whether the evolution trend shows rapid deterioration), and analyzes the degree of deviation of the Ra value from the target roughness specification.

[0093] Based on these analysis results, the system sets processing priorities for various input information and builds a dynamic priority task queue. This creates a processing priority queue. For example, if the system detects high-risk crack development or Ra values ​​that are continuously and severely exceeded, the task will be assigned the highest priority, ensuring that critical process issues are responded to first.

[0094] The rule engine integrates the surface defect information, the defect evolution trend information, the surface roughness information, and the thermal anomaly information based on the processing priority queue, applies an adjustment action rule library, and generates a report on the physical properties of the defect surface.

[0095] The rule base is a structured logic set that can be configured and maintained by process experts and is constructed using an "IF-THEN" logic format. Specifically, it includes:

[0096] Conditional part (IF): consists of the combination logic of one or more detection parameters;

[0097] Action part (THEN): Explicit instructions to perform specific adjustments to one or more processing parameters, such as adjusting the grinding head pressure, spindle speed, feed rate, abrasive concentration or coolant flow rate; the adjustment form includes percentage increase or decrease, setting target value or switching process parameter groups.

[0098] If multiple rules are triggered simultaneously, the system processes them based on preset rule priorities, conflict resolution strategies, or meta-rules. These strategies include: selecting a conservative adjustment plan (to avoid over-response); weighted integration of multiple recommended adjustment ranges; and prioritizing high-risk response actions based on risk levels.

[0099] Through the aforementioned rule matching and execution process, the rule engine ultimately generates a detailed, actionable report on the physical characteristics of the defect surface. These recommendations are output in a standardized data format (e.g., including parameter name, target adjustment value or adjustment range, and execution priority).

[0100] By incorporating a priority queue mechanism, the system prioritizes responses to anomalies that most significantly impact diamond quality, improving the timeliness and effectiveness of feedback. Furthermore, the configurable rule base allows experts to embed their expertise and optimization strategies into the system, and the control logic can be flexibly adjusted based on actual production needs. This enables more refined, intelligent, and efficient closed-loop feedback control of the diamond polishing process, thereby improving final product qualification rates and overall production line efficiency.

[0101] Furthermore, the defect information synthesis unit also includes: using the isolation forest algorithm and the principal component analysis reconstruction error analysis method to compare the surface defect information, the thermal anomaly information and the surface roughness information with the normal operating condition model to identify new surface anomaly states, and the new surface anomaly states are used to trigger and update the rule engine to generate the defect surface physical property report.

[0102] Based on the isolation forest algorithm and principal component analysis reconstruction error analysis method, the online detection system's ability to identify and respond to unknown risks and new failure modes in the diamond polishing process is enhanced. This detection method does not rely on predefined defect types or rules. By comparing with historical normal operating condition models, it can sensitively capture statistically significant abnormal conditions such as surface quality issues caused by sudden equipment problems or unseen diamond defects in real time. Once such new anomalies are identified, not only can early warnings or emergency interventions be triggered, but these abnormal events and their related data also provide valuable input for subsequent updates and optimizations of the rule engine. This enables the system to learn from the unknown, continuously expand its knowledge base and response capabilities, improve the robustness and intelligence of the entire production process, and avoid production losses that may be caused by the failure to promptly identify and address new anomalies.

[0103] This invention encompasses three core components: multimodal data acquisition and preprocessing, multidimensional intelligent analysis and interpretation, and real-time feedback closed-loop control. It enables fully automated quality monitoring and process optimization for the diamond polishing process. The system comprehensively captures workpiece surface condition information, including optically visible defects, thermally induced anomalies, precise roughness, and defect evolution trends. Through intelligent decision-making, the system dynamically adjusts process parameters based on real-time feedback, effectively improving diamond product quality and consistency, increasing production efficiency, and enhancing the intelligent and adaptive capabilities of the production process.

[0104] Example 2:

[0105] To further explain and illustrate the first embodiment in detail, a diamond polishing and grinding online detection system includes:

[0106] Defect data acquisition unit, used to synchronously acquire optical sensing data and thermal sensing data of the diamond surface, and pre-process and output optical characterization data containing surface physical property information;

[0107] Table 1 Comparison of surface defect detection performance

[0108]

[0109] A defect characteristic analysis unit is used to detect defects on the diamond surface based on the optical characterization data by fusing a data-driven model with a visual detection model, output surface defect information, and detect thermal anomalies in combination with the thermal sensing data, and output thermal anomaly information; calculate the surface roughness of the diamond surface based on the optical characterization data, and self-calibrate the surface roughness calculation by comparing it with external reference data, and output surface roughness information; and compare the surface defect information with a historical defect database to output defect evolution trend information.

[0110] Specifically, in order to demonstrate the effectiveness of the present invention in detecting defects by fusing the data-driven model with the visual detection model, the performance of surface defect detection is compared as shown in Table 1. It can be seen that the defect detection effect of the present invention is the best.

[0111] The defect information synthesis unit is used to make feedback decisions based on the surface defect information, the thermal anomaly information, the surface roughness information and the defect evolution trend information using a rule engine, generate a defect surface physical property report, and output it to the controller through an industrial communication interface.

[0112] Furthermore, the defect information synthesis unit also includes: using the isolation forest algorithm and the principal component analysis reconstruction error analysis method to compare the surface defect information, the thermal anomaly information and the surface roughness information with the normal operating condition model to identify new surface anomaly states, and the new surface anomaly states are used to trigger and update the rule engine to generate the defect surface physical property report.

[0113] Specifically, the system first integrates multi-source information from the current inspection cycle from the image feature analysis unit, including surface defect information, thermal anomaly information, and surface roughness information, to construct a feature vector that comprehensively represents the current workpiece surface state. This feature vector is then compared with a normal operating condition model. The normal operating condition model is constructed based on a large number of historical samples of workpiece data that have been confirmed to be in a stable, high-quality state. It can be regarded as a healthy baseline in a multidimensional feature space.

[0114] The system constructs and stores a normal operating condition model or baseline model in advance by learning a large amount of historical detection data (i.e., the set of the above-mentioned comprehensive feature vectors) that is confirmed to represent normal or high-quality production conditions.

[0115] This anomaly detection function uses a technical path that combines the isolation forest algorithm with the principal component analysis (PCA) reconstruction error analysis method to compare the current comprehensive feature vector with the normal operating condition model in real time and calculate its anomaly score or the degree of deviation from the normal pattern.

[0116] If the anomaly score calculated by the isolation forest algorithm exceeds the preset threshold, or the PCA reconstruction error is significantly larger than the normal fluctuation range, the system determines that a new type of surface abnormality exists on the current workpiece surface. This type of anomaly is usually caused by unknown or uncovered process disturbances, such as abnormal wear of the mold, failure of the cooling system, material anomalies, etc., which may cause conventional rules to be unable to respond effectively. The new abnormal state can also be recorded as a key event and used as input to participate in subsequent rule engine updates and expansions. On the one hand, the event can be included in the rule engine rule library through manual annotation or online learning modules to supplement previously uncovered scenarios; on the other hand, the identified abnormal information can also be used to adjust the risk assessment logic and optimize the generation strategy of future defect surface physical property reports. Through the above-mentioned anomaly identification and response mechanism, the system can break through the limitations of the rule engine for known problems and has the ability to identify unknown anomalies, thereby improving the sensitivity and response speed to potential quality risks and enhancing the system's intelligent decision-making capabilities and robustness under complex and dynamic working conditions.

[0117] In order to verify the technical advantages of the defect information synthesis unit of the present invention in terms of intelligence, response speed and adaptability, a comparison with two representative baseline models was constructed, as shown in Table 2. First, in terms of feedback decision response time, it is not much different from the response time of other baseline models, and can meet the real-time requirements of the diamond online production line. Secondly, in terms of abnormality recognition and feedback accuracy, the present invention can still achieve the highest recognition rate when dealing with sudden and non-preset surface conditions (such as cooling failure, grinding tool shedding), while the baseline model is generally unable to effectively deal with unknown abnormalities and has obviously insufficient detection capabilities. In terms of quality results, the present invention can reduce the surface defect rate and roughness fluctuation range, among which the Ra fluctuation standard deviation is reduced the most, while the basic model has limited improvement. Correspondingly, the additional scrap rate caused by improper feedback is the lowest, and the present invention is significantly better than other baseline models.

[0118] Table 2 Comparison of key performance data

[0119]

[0120] 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. A diamond polishing and grinding online detection system, characterized in that: include: Defect data acquisition unit, used to synchronously acquire optical sensing data and thermal sensing data of the diamond surface, and pre-process and output optical characterization data containing surface physical property information; Defect characteristic analysis unit, which is used to detect defects on the diamond surface based on optical characterization data by fusing a data-driven model with a visual detection model, output surface defect information, and detect thermal anomalies in combination with thermal sensing data, and output thermal anomaly information; The surface roughness of the diamond surface is calculated based on the optical characterization data, and the surface roughness calculation is self-calibrated by comparing with the external benchmark data, and the surface roughness information is output. The process of outputting the surface roughness information includes: based on the optical characterization data, applying fast Fourier transform processing to analyze the spectral energy distribution of the low-frequency approximate component and the high-frequency detail component; and estimating the surface roughness and quantifying the texture feature parameters based on the predetermined estimation model between the spectral energy distribution and the surface roughness to obtain preliminary surface roughness information; regularly receiving the benchmark surface roughness reference value for the same sample area through the calibration data interface; and applying the adaptive calibration algorithm to obtain the surface roughness information through comparison. Calculating a deviation between the surface roughness and a reference surface roughness value; adjusting internal parameters of a predetermined estimation model based on the deviation, recalculating the surface roughness, and outputting calibrated surface roughness information; and comparing the surface defect information with a historical defect database to output defect evolution trend information. The process of outputting the defect evolution trend information includes: comprehensively considering the proximity of the defect spatial position, potential motion trajectory prediction, and the spatiotemporal attributes of the corresponding processing stage timestamp, while matching the multi-dimensional characteristics of the defect type, size, and shape factor to identify whether it belongs to an evolution body of the same defect in the historical database, thereby establishing a tracking relationship and obtaining an associated defect number. The defect information synthesis unit is used to make feedback decisions based on surface defect information, thermal anomaly information, surface roughness information and defect evolution trend information using a rule engine, generate a report on the physical properties of the defect surface, and output it to the controller through an industrial communication interface.

2. The diamond polishing and grinding online detection system according to claim 1, characterized in that: The pretreatment specifically includes: Converting the optical sensing data into a single-channel grayscale image; applying a contrast-limited adaptive histogram equalization algorithm to the single-channel grayscale image; Using a guided filtering algorithm to perform noise filtering on the contrast-enhanced single-channel grayscale image; And for the filtered single-channel grayscale image, the region of interest is extracted by template matching technology based on normalized cross-correlation, and multi-level wavelet transform is applied to extract and reconstruct low-frequency approximate components and high-frequency detail components to form the optical characterization data.

3. The diamond polishing and grinding online detection system according to claim 1, characterized in that: The process of outputting surface defect information includes: The data-driven model uses a lightweight convolutional neural network; using a lightweight convolutional neural network model that has been pre-trained for diamond surface defect types, it automatically extracts and identifies surface irregular defects from the input optical characterization data and outputs a first defect detection result, including a defect category, a prediction score with a confidence level greater than a preset threshold, and bounding box coordinates; The visual inspection model integrates the Canny edge detection algorithm and the Hough transform algorithm to identify and locate regular surface defects and obtain a second defect detection result; The first defect detection result and the second defect detection result are fused based on a confidence weighting factor and / or an intersection-over-union ratio between bounding boxes, and the surface defect information is generated through redundancy elimination and conflict resolution.

4. The diamond polishing and grinding online detection system according to claim 1, characterized in that: The process of outputting the defect evolution trend information includes: Correlation processing is performed on the surface defect information and the defect map data stored in the historical defect database in terms of time and space dimensions and matching of multi-dimensional characteristic parameters is performed to obtain an associated defect number; Tracking the time-series quantitative change data corresponding to the defect number in the optical sensing data, and analyzing the time-series quantitative change data using a preset time series analysis model to identify and quantify the evolution pattern corresponding to the defect number; wherein the evolution pattern is a parameter vector that characterizes the dynamic statistical characteristics of the defect size change rate, morphological transformation characteristics, and occurrence frequency; Based on the evolution pattern and a preset risk assessment rule set, the future development trend and potential risk level of each corresponding defect number are determined to obtain the defect evolution trend information.

5. The diamond polishing and grinding online detection system according to claim 1, characterized in that: The specific implementation process of the rule engine includes: Analyzing the severity of the surface defect information and the degree of deviation of the surface roughness information from a preset target specification, and setting a processing priority queue; The rule engine integrates the surface defect information, the defect evolution trend information, the surface roughness information, and the thermal anomaly information based on the processing priority queue, applies an adjustment action rule library, and generates a report on the physical properties of the defect surface.

6. The diamond polishing and grinding online detection system according to claim 1, characterized in that: The defect information synthesis unit also includes: using an isolation forest algorithm and a principal component analysis reconstruction error analysis method to compare the surface defect information, the thermal anomaly information and the surface roughness information with a normal operating condition model to identify new surface anomaly states, and the new surface anomaly states are used to trigger and update the rule engine to generate a report on the physical properties of the defect surface.

Images