Metallic silicon incoming material quality monitoring method
By using edge angle analysis algorithms and dynamic curvature threshold algorithms to detect chipping or cracks in metallic silicon, multimodal and multiscale feature fusion to detect scratches, and infrared thermal imagers to detect pores, this technology solves the problem of difficult identification of micro-defects in existing technologies and achieves high-precision automated detection.
Patent Information
- Application Number
- CN202511478695.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-10-16
- Publication Date
- 2026-01-27
AI Technical Summary
Existing detection methods are ineffective in identifying microscopic defects in metallic silicon, such as microcracks and pores, and insufficient professional skills of operators lead to a decline in detection accuracy.
The system employs edge angle analysis and dynamic curvature threshold algorithms to detect chipped edges or cracks, multimodal and multiscale feature fusion to detect scratches, and infrared thermal imagers to detect pores, achieving non-contact three-dimensional defect detection.
It improves the accuracy and automation of silicon metal detection, reduces human error, and increases work efficiency.
Smart Images
Figure CN121410188A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of metal silicon production and testing technology, specifically to a method for monitoring the quality of incoming metal silicon materials. Background Technology
[0002] Metallic silicon, also known as crystalline silicon or industrial silicon, is primarily used as an additive in non-ferrous alloys. It is a product smelted from quartz and coke in an electric furnace, with silicon comprising approximately 98% of its main component, and the remaining impurities being iron, aluminum, calcium, etc. However, the smelting of metallic silicon is an energy-intensive process, and current methods for detecting it have the following limitations: 1. Limitations of detection technology: Existing detection methods (such as visual inspection and gravimetric method) are difficult to effectively identify micro-defects (such as microcracks and pores) and complex geometric problems in metallic silicon, especially with insufficient sensitivity under high purity requirements; although the gravimetric method has high accuracy, it has poor applicability to samples containing special elements such as fluorine and titanium, which increases the risk of missed detection.
[0003] 2. Insufficient professional competence of operators: Lack of skills and insufficient training among inspection personnel can easily lead to operational errors, resulting in decreased testing accuracy; at the same time, weak quality awareness and low execution rate in self-inspection and mutual inspection processes increase the risk of human error.
[0004] Therefore, in order to solve the aforementioned problems, it is necessary to design a method for monitoring the quality of incoming silicon metal. Summary of the Invention
[0005] This invention addresses the problems existing in the prior art by providing a method for monitoring the quality of incoming metallic silicon.
[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A method for quality control of incoming silicon metal includes the following steps: S1. The raw materials produced are pre-treated at the first inspection station; S2. The pre-treated metallic silicon is transported to the second inspection station for defect detection; Among them, defect detection is the detection of three-dimensional defects in metallic silicon, including microcracks, pores and inclusions; and microcracks include edge chipping or cracks, scratches; S3. For defects that meet the criteria after inspection, transmit them to the final collection location.
[0007] Based on the above technical solution, further, in S1, the first detection station is used to detect the percentage of chemical composition of the conveyed raw material and separate the raw material that conforms to the characteristics of metallic silicon.
[0008] Based on the above technical solution, the separation conditions are further as follows: if the silicon content reaches 98% or more, it meets the characteristics of metallic silicon; raw materials with a purity of less than 98% are transported to an external waste collection device; and metallic silicon that meets the conditions is further transported by conveyor belt to the second detection station.
[0009] Based on the above technical solution, further, in S2, the defect detection process is as follows: Step 1: First, inspect the silicon metal for defects such as chipping or cracking. Step 2: Perform scratch defect detection on silicon metal without chipping or cracking defects; Step 3: Perform porosity defect detection on unscratched silicon metal.
[0010] Based on the above technical solution, further, in step 1, the detection process for defects such as chipped edges or cracks is as follows: Step 11: Using an edge angle analysis algorithm, first extract the edge region of the metallic silicon through morphological operations, and then generate a polygonal contour through edge fitting; Step 12: Then, the dynamic curvature threshold algorithm is used to measure the included angle θ of the line connecting three adjacent points; Step 13: Set the edge chipping or crack characteristic angle to 60°. When the included angle θ < 60° and the curvature change exceeds the set value, trigger the edge chipping or crack alarm.
[0011] Based on the above technical solution, further, the process of step 11 is as follows: Step 111, Preprocessing stage: Adaptive median filtering is used to eliminate high-frequency noise, and the filtering window is dynamically adjusted according to the local gray-level variance; Step 112: Perform edge enhancement based on the processing results of Step 11; Step 113: Perform initial fitting based on the edge region enhanced in Step 112 and output a simplified contour; Step 114: Perform secondary optimization based on the simplified outline of the output to generate a polygonal outline.
[0012] Based on the above technical solution, further, in step 2, the scratch defect detection process is as follows: Step 21: Perform multimodal data acquisition; Step 22: Extract defects based on the collected data; Step 23: Perform multi-scale feature fusion on the extracted defect data, and determine whether scratch defects exist based on the fusion structure.
[0013] Based on the above technical solution, further, in step 23, the multi-scale features include at least morphological features and depth features.
[0014] Based on the above technical solution, further, in step 3, the detection process for porosity defects is as follows: Step 31: Use an infrared thermal imager to collect the surface temperature of metallic silicon and capture areas with abnormal thermal diffusion; Step 32: Extract multi-source features from the captured abnormal regions; Step 33: Perform threshold judgment based on the extracted multi-source features. If the set conditions are met, it is considered that there is a porosity defect; otherwise, it is not present.
[0015] Based on the above technical solution, further, in S3, the set compliance conditions are: if at least any 3 defect detections are not present or within an acceptable range, then it is considered to be a compliant silicon metal and is transported to the final collection location; other non-compliant ones are transported to the waste collection device.
[0016] Compared with the prior art, the present invention has the following beneficial effects: This invention can detect three-dimensional defects in metallic silicon from multiple perspectives, including edge chipping or cracking, scratches, and porosity. Specifically, it employs edge angle analysis algorithms and dynamic curvature threshold algorithms to detect edge chipping / scratching defects in metallic silicon; it uses multi-modal, multi-scale feature fusion to detect the presence of scratch defects; and it uses multi-source features to determine the presence of porosity defects. This invention uses non-contact detection of three-dimensional defects, effectively improving detection accuracy. Furthermore, this invention has a high degree of automated monitoring, significantly reducing the error rate of manual operation. Attached Figure Description
[0017] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation
[0018] The present invention will be further described and illustrated below with reference to the accompanying drawings and specific embodiments. The technical features of each embodiment of the present invention can be combined accordingly, provided that there is no mutual conflict.
[0019] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Many specific details are set forth in the following description to provide a thorough understanding of the present invention. However, the present invention can be practiced in many other ways different from those described herein, and those skilled in the art can make similar modifications without departing from the spirit of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below. Technical features in the various embodiments of the present invention can be combined accordingly without mutual conflict.
[0020] Example Combination Figure 1As shown, this embodiment provides a method for monitoring the quality of incoming silicon metal, which includes the following steps: S1. The raw materials produced are pre-treated at the first inspection station; In this embodiment, the produced raw materials are placed on a conveyor belt and conveyed to the first detection station. This first detection station is used to detect the percentage of chemical composition of the conveyed raw materials and separate raw materials that meet the characteristics of metallic silicon. It should be noted that the raw materials are materials produced in the production workshop, including metallic silicon. When separating metallic silicon from the raw materials, the judgment condition is based on the chemical composition of the metallic silicon. It is generally understood that the main component of metallic silicon is silicon element, typically with a purity of 98% or higher. In some embodiments, the silicon content can be determined using professional chemical analysis methods, such as spectroscopic analysis (e.g., inductively coupled plasma mass spectrometry (ICP-MS) or X-ray fluorescence spectrometry (XRF)). If the silicon content reaches 98% or higher, it meets the main component characteristics of metallic silicon. Raw materials with a purity below 98% are conveyed to an external waste collection device. Metallic silicon that meets the criteria continues to be conveyed to the second detection station. S2. The pre-treated metallic silicon is transported to the second inspection station for defect detection; In this embodiment, defect detection mainly detects three-dimensional defects in metallic silicon, including microcracks, pores, and inclusions. Among these, microcracks include chipped edges, cracks, and scratches.
[0021] In this embodiment, the specific detection process is as follows: Step 1: First, inspect the silicon metal for defects such as chipping or cracking. Specifically, the inspection process for defects such as chipped edges or cracks is as follows: Step 11: Using an edge angle analysis algorithm, first extract the edge region of the metallic silicon through morphological operations, and then generate a polygonal contour through edge fitting.
[0022] Among them, the edge angle analysis algorithm is a calculation method used to quantify the distribution of edge directions in an image. It mainly achieves a refined description of edge features through gradient calculation and direction statistics. In other words, the edge angle analysis algorithm extracts image structural features by quantifying the distribution of edge directions.
[0023] More specifically, the process of extraction through morphological operations is as follows: Step 111, Preprocessing stage: Adaptive median filtering is used to eliminate high-frequency noise. The filtering window is dynamically adjusted according to the local gray-level variance. The preferred window size is 3×3 to 7×7. Step 112: Based on the processing results of Step 11, perform edge enhancement: Combine top cap transformation (highlighting bright edges) and bottom cap transformation (highlighting dark edges), and use elliptical structural elements (major-minor axis ratio 2:1) to adapt to irregular chipped edges or cracks.
[0024] This process provides continuous, low-noise edge features for subsequent polygon contour fitting.
[0025] Furthermore, the edge fitting to generate polygonal contours employs a hierarchical contour fitting method, the specific process of which is as follows: Step 113: Perform initial fitting based on the edge region enhanced in Step 112 and output a simplified contour: Compress the edge points of the edge region using the Douglas-Peucker algorithm while preserving macroscopic features. It should be noted that the Douglas-Peucker algorithm (DP algorithm) removes redundant edge points while retaining the global geometric features of the contour (such as straight line segments, large-radius curves, etc.), i.e., preserving macroscopic features; the output is a simplified contour. Specifically, when using the DP algorithm, the maximum allowable fitting error threshold is 0.5% of the contour perimeter. That is, when the farthest distance from the original edge point to the simplified polygon is ≤ 0.5% of the total contour perimeter, the point will be removed.
[0026] Step 114: Perform secondary optimization based on the simplified output contour to generate a polygonal contour: Use Frechet distance detection to insert key points in the curvature abrupt change region (potential defect site) to improve local fitting accuracy. The main reasons for choosing Frechet distance are: (1) it reflects curve similarity better than Euclidean distance (suitable for capturing local distortion of cracks); (2) by calculating the Frechet distance between the original edge and the fitted polygon, the region with the largest positioning deviation (i.e., curvature abrupt change point) is selected according to the situation. It should be noted that these curvature abrupt change points are connected to form the curvature abrupt change region.
[0027] This process is used to ensure that subsequent curvature analysis can both cover the global profile and capture microscopic defects.
[0028] For example, when there is a 2px wide microcrack at the edge of the metal silicon: the DP algorithm may simplify it into a single straight line segment (macro features are preserved); the Frechet distance is used to detect that the matching deviation of this segment from the original edge is >3px, triggering key point insertion; the final fitted profile increases the point density at the crack by 5 times, ensuring that curvature analysis (step 12) can capture the defect.
[0029] Step 12: Next, the dynamic curvature threshold algorithm is used to measure the angle θ between the lines connecting three adjacent points. It should be noted that, based on the output of the DP algorithm in step 11, three consecutive adjacent key points are selected along the polygon contour in a clockwise / counterclockwise direction, such as the two endpoints and inflection point of a straight line segment, or the two endpoints and concave point of an arc segment.
[0030] Among them, the dynamic curvature thresholding algorithm is an adaptive method for image feature detection. By combining curvature calculation with dynamic threshold adjustment, it achieves highly robust corner / feature point extraction.
[0031] Furthermore, the dynamic curvature threshold algorithm process is as follows: Contour extraction and smoothing are performed: ① Use Canny edge detection to obtain the initial contour and fill the gaps (Gap-size ≤ 1 pixel); ② Gaussian filtering is used to smooth the contours: L(x,y)=I(x,y)∗G(x,y,σ); where I(x,y) represents the pixel value of the input image at coordinates (x,y), which is the original image data to be processed; G(x,y,σ) is the Gaussian kernel function, which implements the weighted average filter; σ is the control scale, used to control the smoothing intensity; L(x,y) represents the pixel value of the output image at coordinates (x,y), which is the smoothed image data.
[0032] Curvature calculation and scale space construction are performed, and a curvature-scale space map is generated. Specifically, the first / second derivatives of the contour points are calculated, and the curvature K is calculated in the multi-scale space to generate a curvature-scale space map. Generate dynamic thresholds; ① For each contour point, calculate local statistics by taking its neighborhood window (e.g., radius r = 10 pixels); ② If the curvature K at the current point is a local maximum, mark it as a candidate corner point; Pseudo-corner filtering; ① Rounded corner filtering: Removes points with excessively large curvature circle radius R=1 / K; ②Endpoint processing: Determine whether to retain the contour endpoints based on the flag bit (retained by default).
[0033] Step 13: Set the edge chipping or crack characteristic angle to 60°. When the included angle θ < 60° and the curvature abrupt change exceeds the set value (e.g., 0.15 rad / mm), trigger the edge chipping or crack alarm.
[0034] This detection process solves the problem of missed detection of chamfered edges by traditional image methods, and can effectively improve the accuracy of precision detection.
[0035] Step 2: Perform scratch defect detection on silicon metal without chipping or cracking defects; Specifically, the inspection process for scratches is as follows: Step 21: Perform multimodal data acquisition; wherein, a high-precision integrated laser triangulation range sensor can be used to acquire optical image data of the three-dimensional depth of the scratch in real time; this step 21 provides a complementary physical feature dimension for step 22; Step 22: Defect extraction based on the collected data; specifically, anisotropic diffusion filtering is applied to the optical image to preserve edge details, then the surface gradient field is calculated and a curvature map is generated, and the threshold is dynamically adjusted based on the curvature map statistics to extract candidate scratch regions (curvature extrema points), wherein the curvature extrema points are set according to the actual situation; Step 23: Perform multi-scale feature fusion on the extracted defect data, and determine whether scratch defects exist based on the fusion structure; Furthermore, this multi-scale feature includes at least morphological and depth features. The morphological feature can be extracted using skeletonization and fractal dimension analysis to effectively distinguish scratches from particulate impurities. The depth feature can be extracted using Gaussian fitting of light-gathering point clouds to effectively quantify the aspect ratio of the scratch. A scratch defect is considered to exist when all the set scale features are satisfied; otherwise, no scratch defect exists.
[0036] Step 3: Perform porosity defect detection on unscratched silicon metal; Specifically, the detection process for pore defects is as follows: Step 31: Use an infrared thermal imager to collect the surface temperature of the silicon metal and capture areas with abnormal thermal diffusion. It should be noted that a threshold range for the surface temperature of the silicon metal is set. When the temperature of thermal diffusion is outside the threshold range, it is considered an abnormal area. Step 32: Perform multi-source feature extraction on the captured abnormal region; wherein, the multi-source feature extraction includes morphological feature extraction and physical feature extraction; further, the morphological features include sphericity and aspect ratio of pores; the physical features include thermal diffusion delay time and sound wave scattering angle.
[0037] Step 33: Threshold judgment is performed based on the extracted multi-source features. If the set conditions are met, it is considered that a porosity defect exists; otherwise, it does not. It should be noted that the threshold ranges for porosity sphericity, aspect ratio, thermal diffusion delay time, and sound wave scattering angle are all set according to actual needs.
[0038] S3. For those that meet the defect detection criteria, they are transferred to the final collection location. It should be noted that the qualification criteria are: if at least any three defect detections are not present or within acceptable limits, then it is considered a qualified piece of metallic silicon and is transferred to the final collection location; other pieces that do not meet the criteria are sent to the waste collection device.
[0039] In this embodiment, a non-contact detection process is mainly adopted, which improves the degree of automation and accuracy of detection, reduces the error rate and labor cost of manual detection, and thus improves work efficiency.
[0040] Finally, it should be noted that the above content is only used to illustrate the technical solution of the present invention, and is not intended to limit the scope of protection of the present invention. Simple modifications or equivalent substitutions made by those skilled in the art to the technical solution of the present invention do not depart from the essence and scope of the technical solution of the present invention.
Claims
1. A method for quality control of incoming metallic silicon, characterized in that, Includes the following steps: S1. The raw materials produced are pre-treated at the first inspection station; S2. The pre-treated metallic silicon is transported to the second inspection station for defect detection; Among them, defect detection is the detection of three-dimensional defects in metallic silicon, including microcracks, pores and inclusions; and microcracks include edge chipping or cracks, scratches; S3. For defects that meet the criteria after inspection, transmit them to the final collection location.
2. The method for monitoring the quality of incoming metallic silicon according to claim 1, characterized in that, In S1, the first detection station is used to detect the percentage of chemical composition of the conveyed raw materials and separate the raw materials that conform to the characteristics of metallic silicon.
3. The method for monitoring the quality of incoming metallic silicon according to claim 2, characterized in that, The separation conditions are as follows: if the silicon content reaches 98% or more, it meets the characteristics of metallic silicon; raw materials with a purity of less than 98% are transported to an external waste collection device; while metallic silicon that meets the conditions continues to be transported by conveyor belt to the second testing station.
4. The method for monitoring the quality of incoming metallic silicon according to claim 1, characterized in that, In S2, the defect detection process is as follows: Step 1: First, inspect the silicon metal for defects such as chipping or cracking. Step 2: Perform scratch defect detection on silicon metal without chipping or cracking defects; Step 3: Perform porosity defect detection on unscratched silicon metal.
5. The method for monitoring the quality of incoming metallic silicon according to claim 4, characterized in that, In step 1, the inspection process for defects such as chipped edges or cracks is as follows: Step 11: Using an edge angle analysis algorithm, first extract the edge region of the metallic silicon through morphological operations, and then generate a polygonal contour through edge fitting; Step 12: Then, the dynamic curvature threshold algorithm is used to measure the included angle θ of the line connecting three adjacent points; Step 13: Set the edge chipping or crack characteristic angle to 60°. When the included angle θ < 60° and the curvature abrupt change exceeds the set value, trigger the edge chipping or crack alarm.
6. The method for monitoring the quality of incoming metallic silicon according to claim 5, characterized in that, The process of step 11 is as follows: Step 111, Preprocessing stage: Adaptive median filtering is used to eliminate high-frequency noise, and the filtering window is dynamically adjusted according to the local gray-level variance; Step 112: Perform edge enhancement based on the processing results of Step 11; Step 113: Perform initial fitting based on the edge region enhanced in Step 112 and output a simplified contour; Step 114: Perform secondary optimization based on the simplified outline of the output to generate a polygonal outline.
7. The method for monitoring the quality of incoming metallic silicon according to claim 4, characterized in that, In step 2, the process for detecting scratch defects is as follows: Step 21: Perform multimodal data acquisition; Step 22: Extract defects based on the collected data; Step 23: Perform multi-scale feature fusion on the extracted defect data, and determine whether scratch defects exist based on the fusion structure.
8. The method for monitoring the quality of incoming metallic silicon according to claim 7, characterized in that, In step 23, the multi-scale features include at least morphological features and depth features.
9. The method for monitoring the quality of incoming metallic silicon according to claim 4, characterized in that, Step 3, the detection process for porosity defects is as follows: Step 31: Use an infrared thermal imager to collect the surface temperature of metallic silicon and capture areas with abnormal thermal diffusion; Step 32: Extract multi-source features from the captured abnormal regions; Step 33: Perform threshold judgment based on the extracted multi-source features. If the set conditions are met, it is considered that there is a porosity defect; otherwise, it is not present.
10. The method for monitoring the quality of incoming metallic silicon according to claim 1, characterized in that, In S3, the compliance conditions are set as follows: if at least any 3 defect detections are not present, it is considered to be a compliant silicon metal and is transported to the final collection location; other non-compliant materials are transported to the waste collection device.