Quality detection method and system for aluminum alloy polishing

Abstract

The invention relates to the technical field of aluminum alloy polishing quality detection, and discloses a quality detection method and system for aluminum alloy polishing, and the method comprises the steps: obtaining a preliminary surface feature data set, and carrying out the denoising processing to obtain surface feature refining data; decomposing the data into related components of flatness and roughness; according to the flatness related component, geometric deviation calculation is carried out to obtain a flatness anomaly distribution diagram; in combination with the roughness correlation component, performing local region comparative analysis to obtain a roughness anomaly correlation result; the two types of data are fused for comprehensive correlation analysis, and a surface quality comprehensive feature matrix is obtained; performing weighted integration on the characteristics of the flatness and the roughness to obtain a quality evaluation distribution diagram; and generating detection feedback data according to the quality evaluation distribution diagram. According to the method, the flatness and the roughness of the aluminum alloy surface can be synchronously detected, and the accuracy and the real-time performance of polishing quality evaluation are improved.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of quality detection of aluminum alloy polishing, and in particular to a quality detection method and system for aluminum alloy polishing. BACKGROUND

[0002] Aluminum alloy refers to an alloy taking aluminum as a base and adding a certain amount of other alloying elements, and the aluminum alloy is divided into wrought aluminum alloy and cast aluminum alloy according to its composition and processing method; wherein, the wrought aluminum alloy is first melted into an ingot, and then plastic deformation processing is performed, and various plastic processing products are prepared through rolling, extrusion, stretching, forging and other methods; and the cast aluminum alloy refers to a blank that is directly cast into various parts by using sand mold, iron mold, investment mold and die casting method after the ingredients are melted; and no matter the cast aluminum alloy or the wrought aluminum alloy, the surface of the parts or products is polished and ground after being processed into parts or products, so as to ensure the smoothness of the surface of the parts or products, so as to prevent the parts or products from affecting the user experience effect during use.

[0003] The aluminum alloy material is widely used in electronic manufacturing, automobile industry and other manufacturing fields due to its excellent strength-to-weight ratio and good processing performance; for example, in the processing process of aluminum alloy plate, the cylindrical aluminum rod is usually heated and melted first, then other alloying elements are added into the molten aluminum water, and the aluminum alloy solution is cooled to form a plate-shaped aluminum alloy plate (for subsequent processing and transportation), the aluminum alloy plate is cut and processed into target parts or products, and the formed aluminum alloy plate, parts or products are polished to obtain the target product; at present, as the basic material, how to determine the quality of the surface of the aluminum alloy plate after polishing and grinding is a key link in the production process when the surface of the aluminum alloy plate is polished and ground; if the quality of the surface of the aluminum alloy plate is not good (such as having protrusions, aluminum scraps, burrs, concave points and other defects), it will extremely easily cause quality problems and safety problems in the subsequent processing link, such as: the edge of the aluminum alloy plate cutting the worker, the rough surface of the aluminum alloy plate damaging the parts of the equipment and the poor surface quality of the aluminum alloy product; therefore, the quality of the aluminum alloy plate not only directly affects the appearance of the product, but also relates to the smooth progress of the subsequent processing and the service life; therefore, the detection of the polishing quality of the aluminum alloy has become a very important technical direction in the manufacturing industry, and has important significance for improving the overall quality of industrial products and market competitiveness.

[0004] In the prior art, the polishing quality is usually evaluated by measuring the flatness and roughness of the surface of the aluminum alloy, or the surface quality of the aluminum alloy plate is determined by manually observing whether the surface is smooth. However, the former detection method is difficult to capture subtle topographic changes when processing complex curved surfaces or large-sized workpieces due to the lack of high-performance intelligent sensing support in the detection system, especially in the accurate acquisition of flatness information. The latter detection method is prone to reduce the detection efficiency and accuracy, and may lead to rework of the aluminum alloy plate, thereby increasing the manufacturing cost. In addition, the flatness and roughness are highly correlated, and the error of the flatness will further affect the judgment of the surface roughness, resulting in that the detection result cannot truly reflect the actual polishing quality.

[0005] In summary, the prior art cannot realize the synchronous dynamic detection of the flatness and roughness of the surface of the aluminum alloy during the polishing process, and cannot construct a comprehensive quality evaluation system, resulting in deviation between the detection result and the actual polishing effect. Therefore, how to intelligently detect the surface quality of the aluminum alloy plate is a technical problem to be solved by technical personnel. SUMMARY

[0006] The present application provides a quality detection method and system for aluminum alloy polishing to solve the problem of how to detect the surface quality of the aluminum alloy plate.

[0007] In a first aspect, to solve the above technical problems, the present application provides a quality detection method for aluminum alloy polishing, comprising: obtaining a preliminary surface feature data set; performing denoising processing on the preliminary surface feature data set to obtain surface feature refined data; performing hierarchical analysis on the surface feature refined data to extract flatness-related components and roughness-related components; performing geometric deviation calculation according to the flatness-related components, and if the deviation value exceeds a preset flatness threshold, marking as an abnormal region to obtain a flatness abnormal distribution map; performing local region comparison analysis according to the flatness abnormal distribution map and the roughness-related components to obtain a roughness abnormal association result; performing comprehensive correlation analysis by fusing the flatness abnormal distribution map and the roughness abnormal association result to obtain a surface quality comprehensive feature matrix; performing weighted integration on the flatness-related components and the roughness-related components according to the surface quality comprehensive feature matrix, and if the integrated feature value deviates from a preset quality evaluation system standard, marking as a quality substandard region to obtain a quality evaluation distribution map; According to the quality evaluation distribution map, a feature parameter of the quality substandard area is extracted, and detection feedback data is generated.

[0008] In an optional implementation, the obtaining of the preliminary surface feature data set comprises: obtaining three-dimensional topographic data of the aluminum alloy surface, i.e., first topographic data; performing denoising and filtering processing on the first topographic data to obtain second topographic data; performing grid processing according to the second topographic data to obtain third topographic data; performing surface roughness quantitative analysis and feature extraction on the third topographic data to obtain a preliminary surface feature data set.

[0009] In an optional implementation, the denoising processing according to the preliminary surface feature data set to obtain surface feature refined data comprises: performing layer-by-layer separation on surface noise in the preliminary surface feature data set, extracting less disturbed point cloud information, and obtaining optimized point cloud data; if there is data in the optimized point cloud data that exceeds a preset fluctuation threshold range, performing interpolation correction on the data to obtain a smoothed point cloud set; performing anomaly detection on the smoothed point cloud set, and if it is found that a data point has a large deviation from the overall surface details, performing local optimization to obtain feature data conforming to topographic accuracy; performing structured processing on the feature data, converting discrete point cloud data into a structured grid form, and obtaining optimized surface feature refined data.

[0010] In an optional implementation, the surface feature refined data is decomposed to extract flatness-related components and roughness-related components, comprising: performing hierarchical analysis on the surface feature refined data to obtain flatness components and roughness components, and integrating to obtain a first decomposition data set; performing feature correlation mapping processing on the flatness components and the roughness components according to the first decomposition data set, detecting fluctuation of surface details, and if it is detected that surface details in the components exceed a preset detail fluctuation threshold range, performing local adjustment and correction to obtain a second decomposition data set; performing quantitative processing on the flatness components and the roughness components according to the second decomposition data set, and extracting surface flatness and surface roughness, and if the surface flatness and the surface roughness do not conform to preset flatness and roughness thresholds, corresponding data is deleted to obtain a feature parameter set; performing classification and arrangement on the feature parameter set to obtain final flatness-related components and roughness-related components.

[0011] In an optional implementation, the geometry deviation calculation according to the flatness-related component is performed, and if the deviation value exceeds a preset flatness threshold, the area is marked as an abnormal area to obtain a flatness abnormality distribution map, including: performing deviation calculation on the flatness-related component to obtain a flatness deviation value; If the flatness deviation value exceeds a preset flatness threshold range, the area exceeding the threshold is marked as an abnormal area to obtain an abnormal area set; According to the abnormal area set, the abnormal area is subjected to a detailed hierarchical processing to obtain hierarchical area data; Combine the hierarchical area data with a set of preset standard parameters to perform visual processing to generate a flatness abnormality distribution map.

[0012] In an optional implementation, the local area comparison analysis is performed according to the flatness abnormality distribution map and the roughness-related component to obtain a roughness abnormality correlation result, including: According to the flatness abnormality distribution map, the abnormal area is preliminarily divided to obtain key area position information; From the roughness-related component, the roughness-related component corresponding to the key area position information is extracted and mapped one by one to obtain a target interval with a larger parameter fluctuation amplitude; Correlate the target interval with the change trend feature to obtain the trend distribution of the roughness parameter; If the trend distribution exceeds a preset roughness threshold range, the trend distribution is fused with the hierarchical area data to obtain a roughness abnormality correlation result.

[0013] In an optional implementation, the flatness abnormality distribution map and the roughness abnormality correlation result are fused to perform comprehensive correlation analysis to obtain a surface quality comprehensive feature matrix, including: According to the flatness abnormality distribution map and the roughness abnormality correlation result, the abnormal area is divided, and the distribution trend information in the area is extracted, the correlation between the distribution trend information and a preset quality index is determined to determine the key position data of the abnormal area; According to the key position data, the roughness-related component and the flatness abnormality distribution map are dynamically mapped and analyzed to obtain a potential fluctuation interval of the surface quality; If the fluctuation interval exceeds a preset quality fluctuation threshold range, the roughness-related component and the flatness abnormality distribution map are integrated to generate a comprehensive feature data set; According to the comprehensive feature data set and the distribution trend information, a surface quality comprehensive feature matrix is constructed.

[0014] In an optional implementation, according to the surface quality comprehensive feature matrix, the features of flatness and roughness are weighted and integrated, if the integrated feature value deviates from a preset quality evaluation system standard, a quality substandard area is marked, and a quality evaluation distribution map is obtained, including: obtaining a surface quality overall distribution map; According to the surface quality comprehensive feature matrix, the feature weights of the flatness related component and the roughness related component are calculated one by one to obtain a first feature set after weighted integration, and the first feature set is compared with a preset quality standard to obtain a comparison result of the first feature set; If the comparison result of the first feature set exceeds a preset comparison threshold range, the substandard area in the first feature set is preliminarily marked to obtain a marked area data set; According to the area data set, the substandard area is associated with the overall distribution map of the surface quality for correlation processing, abnormal position information in the distribution map is extracted, whether the abnormal position information conforms to a preset quality evaluation system standard is judged, and a judgment result of the abnormal position information is obtained; According to the judgment result of the abnormal position information, the distribution map is finally adjusted to generate a final quality evaluation distribution map containing a substandard area mark.

[0015] In an optional implementation, according to the quality evaluation distribution map, the feature parameters of the quality substandard area are extracted to generate detection feedback data, including: obtaining a historical adjustment record of a polishing process; According to the quality evaluation distribution map, the quality substandard area is subjected to separation processing of feature parameters, surface roughness values and flatness values are extracted, and a preliminary data set is obtained; The preliminary data set is matched with a preset polishing process parameter library, if the surface roughness values and flatness values in the preliminary data set exceed a preset matching threshold range, the quality substandard area is classified and marked, and a classified feature parameter set is determined; The classified feature parameter set is associated with a preset adjustment scheme database of the polishing process to obtain preliminary feedback data for the quality substandard area; According to the preliminary feedback data, the preliminary feedback data is fused with the historical adjustment record of the polishing process to generate final detection feedback data.

[0016] In a second aspect, the present application provides a quality detection system for aluminum alloy polishing, comprising: a data acquisition module configured to acquire a preliminary surface feature data set; a data denoising module configured to perform denoising processing on the preliminary surface feature data set to obtain surface feature refined data; a feature decomposition module configured to perform hierarchical analysis on the surface feature refined data to extract a flatness-related component and a roughness-related component; a flatness anomaly detection module configured to perform geometric deviation calculation according to the flatness-related component, and if the deviation value exceeds a preset flatness threshold range, mark the area as an abnormal area to obtain a flatness anomaly distribution map; a roughness anomaly analysis module configured to perform local area comparative analysis according to the flatness anomaly distribution map and the roughness-related component to obtain a roughness anomaly correlation result; a feature fusion module configured to perform comprehensive correlation analysis on two types of data according to the flatness anomaly distribution map and the roughness anomaly correlation result to obtain a surface quality comprehensive feature matrix; a quality evaluation module configured to perform weighted integration on the flatness and roughness features according to the surface quality comprehensive feature matrix, and if the integrated feature value deviates from a preset quality evaluation system standard, mark the area as a quality substandard area to obtain a quality evaluation distribution map; a feedback generation module configured to extract feature parameters of the quality substandard area according to the quality evaluation distribution map to generate detection feedback data.

[0017] In a third aspect, the present application further provides an electronic device comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the quality detection method of the aluminum alloy polishing.

[0018] In a fourth aspect, the present application further provides a computer readable storage medium comprising a stored computer program, wherein the computer program controls a device where the computer readable storage medium is located to execute the quality detection method of the aluminum alloy polishing when the computer program is running.

[0019] Compared with the prior art, the present application has the following beneficial effects: (1) The application carries out high-resolution scanning on the surface of the aluminum alloy through the optical imaging system, generates initial three-dimensional topographic data for complex curved surface topography, covers the characteristic information of surface flatness and roughness, obtains a preliminary surface feature data set, and in this process, layered processing reduces the data processing burden, denoising and filtering improve the data reliability, and gridding processing provides an intuitive model for roughness analysis. These steps work together to not only improve the analysis efficiency, but also provide a scientific basis for subsequent surface quality evaluation, significantly reducing the risk of misjudgment caused by data errors.

[0020] (2) In the process of denoising the preliminary surface feature data set, the application uses data filtering and layer-by-layer separation technology to identify and eliminate environmental interference factors, and combines interpolation correction and anomaly detection to smooth and optimize local noise points, ensuring the continuity and accuracy of the point cloud data. At the same time, the optimized point cloud is structured into a triangular mesh model, improving the consistency and topography restoration of the data, and finally forming a feature refined data set, providing a reliable data foundation and stable model support for high-precision surface quality analysis.

[0021] (3) When processing the surface feature refined data, the application uses layered analysis to decompose it into flat components and rough components, identifies abnormal fluctuation areas through layer-by-layer stripping and feature mapping, and adjusts the local data, effectively improving the accuracy of the component data. Then through parameter extraction and classification, a clear structure and quantifiable feature parameter data set is formed. This process enhances the recognition ability of surface detail features, improves the depth and stability of data analysis under complex topography, and provides high-quality data support for subsequent quality evaluation.

[0022] (4) When analyzing the deviation of the flatness-related components, the application accurately calculates the deviation value by layer-by-layer analyzing the geometric features of the data and comparing it with the preset threshold to identify abnormal areas, and improves the positioning accuracy of abnormal identification by finely layering the abnormal areas. The deviation data is visualized to form an intuitive flatness abnormality distribution map, making it easier to identify and evaluate abnormal information, and providing a scientific and detailed basis for subsequent defect analysis and surface repair.

[0023] (5) Based on the flatness abnormal area, the application introduces the roughness parameter for local area comparative analysis, realizes the accurate identification of the roughness change trend by extracting the key area and comparing the roughness fluctuation amplitude, and fuses the roughness abnormality and area layering information to generate roughness abnormality correlation distribution data, which not only reveals the change characteristics of the abnormal area, but also provides data support for comprehensive evaluation of surface quality, effectively improving the depth and reliability of abnormal detection.

[0024] (6) Based on roughness and smoothness anomaly data, this invention uses parameter comparison and data fusion methods to construct dynamic mapping relationship, realize multi-dimensional feature integration of key areas of surface quality, extract basic information through comprehensive feature dataset, and generate comprehensive surface quality feature matrix by combining regional division. This not only reveals the correlation between different indicators, but also systematically improves the accuracy and completeness of surface anomaly identification, and provides a quantitative basis for the quality assessment of complex morphology.

[0025] (7) Based on the surface quality comprehensive feature matrix, the present invention integrates the flatness and roughness features with weights, calculates the first feature set, compares it with the quality assessment system standard, identifies the deviation area and performs visualization marking, and finally generates a quality assessment distribution map containing abnormal markers. This not only realizes the unified evaluation of multiple quality features, but also improves the accuracy of surface defect location and the pertinence of treatment.

[0026] (8) Based on the final quality assessment distribution map, the present invention extracts the roughness and flatness feature parameters of the substandard areas, performs matching analysis with the polishing process parameter library, classifies and marks the substandard areas, maps their parameter sets with the adjustment scheme database, and generates targeted detection feedback data by combining historical process adjustment records, thereby providing clear and executable improvement references for the optimization of polishing process and improving the accuracy and dynamic adaptability of surface quality control. Attached Figure Description

[0027] Figure 1 This is a flowchart illustrating a quality inspection method for aluminum alloy polishing provided in the first embodiment of the present invention. Figure 2 This is a schematic diagram of a quality inspection system for aluminum alloy polishing provided in the second embodiment of the present invention. Detailed Implementation

[0028] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0029] Reference Figure 1 The first embodiment of the present invention provides a quality inspection method for aluminum alloy polishing, comprising the following steps: S1, Obtain the preliminary surface feature dataset; S2, Denoise the preliminary surface feature dataset to obtain refined surface feature data; S3, performing hierarchical analysis on the surface feature refinement data to extract flatness-related components and roughness-related components; S4, performing geometric deviation calculation according to the flatness-related components, and if the deviation value exceeds a preset flatness threshold, marking it as an abnormal area to obtain a flatness abnormality distribution map; S5, performing local area comparative analysis according to the flatness abnormality distribution map and the roughness-related components to obtain roughness abnormality correlation results; S6, fusing the flatness abnormality distribution map and the roughness abnormality correlation results to perform comprehensive correlation analysis to obtain a surface quality comprehensive feature matrix; S7, according to the surface quality comprehensive feature matrix, weighting and integrating the flatness-related components and the roughness-related components, and if the integrated feature value deviates from a preset quality evaluation system standard, marking it as a quality substandard area to obtain a quality evaluation distribution map; S8, according to the quality evaluation distribution map, extracting feature parameters of the quality substandard area to generate detection feedback data.

[0030] In step S1, a preliminary surface feature data set is obtained, including: Obtaining three-dimensional topography data of the aluminum alloy surface, i.e., obtaining first topography data Performing denoising and filtering processing on the first topography data to obtain second topography data; Performing grid processing on the second topography data to obtain third topography data; Performing surface roughness quantification analysis and feature extraction on the third topography data to obtain a preliminary surface feature data set.

[0031] It should be noted that the aluminum alloy surface is scanned by a high-precision optical scanning device to obtain three-dimensional topography data of a complex curved surface, and the first topography data is obtained. Taking a scanning area of 100 square centimeters and a resolution of 0.01 millimeters as an example, point cloud data containing millions of points can be collected. Such high-resolution collection can provide detailed basic data for subsequent analysis.

[0032] Subsequently, noise points generated due to light interference during the scanning process are removed by filtering to generate a second topography data set. In one possible implementation, if 1% of the point cloud data in the original data is noise, the noise ratio can be reduced to 0.1% after filtering, significantly improving the data quality. For outliers in the point cloud, if the height value of a certain point deviates from the average value by more than a preset threshold such as 0.5 millimeters, the correct value of the point is estimated by the values in the domain (radius 0.5 cm), and the outlier is corrected, thereby generating a more accurate surface feature set, ensuring that the refined surface feature set is closer to the true topography.

[0033] After denoising and outlier correction, the second topography dataset is gridded to generate a third topography dataset. Preferably, the point cloud data is converted into a curved surface grid structure. For example, a triangular mesh is constructed with a grid density of 0.02 millimeters, which can more realistically restore the complex geometric topography of the aluminum alloy surface and provide an intuitive and visual three-dimensional model for subsequent analysis.

[0034] When performing surface roughness quantification analysis based on the third topography dataset, the topography data is reconstructed in three dimensions, a specific area is intercepted, and the line profile analysis method or the area average method is used to calculate the surface roughness parameters, such as the arithmetic average roughness Ra value. Assuming that the measured value is 1.2 microns, this data can be used to evaluate the processing quality of the aluminum alloy surface, and then generate the final roughness feature dataset. Such quantitative analysis helps to determine whether the surface meets the industrial standards, for example, whether it is suitable for the manufacture of aerospace components.

[0035] In step S2, the preliminary surface feature dataset is denoised to obtain surface feature refined data, including: The surface noise in the preliminary surface feature dataset is separated layer by layer to extract point cloud information with less interference to obtain optimized point cloud data; If there is data in the optimized point cloud data that exceeds the preset fluctuation threshold range, the data is corrected by interpolation to obtain a smoothed point cloud set; The smoothed point cloud set is subjected to anomaly detection, and if the data points deviate greatly from the overall surface details, local optimization is performed to obtain feature data that meets the topography accuracy; The feature data is structured to convert the discrete point cloud data into a structured grid form to obtain optimized surface feature refined data.

[0036] Specifically, first, by analyzing the spatial distribution characteristics of the point cloud data, abnormal data fluctuations introduced by factors such as light changes and equipment jitter are identified as noise, and point cloud information with less interference is separated layer by layer to form optimized point cloud data. In one possible implementation, the layer-by-layer separation method is as follows: the point cloud data obtained by scanning the aluminum alloy surface can be divided into multiple intervals according to the depth direction, such as 0-3 millimeters, 3-6 millimeters, etc., and the noise points in each interval are processed respectively. This layer-by-layer separation method can more accurately preserve the true topography information of the surface.

[0037] Further, for the optimized point cloud data, by detecting the change of the height value of the point cloud in the local area, it is judged whether the remaining noise points exist fluctuation exceeding the preset threshold. For example, if the set fluctuation threshold is 0.3mm, for the point cloud data whose height change exceeds the value, the system can correct based on the average height value of multiple points in the neighborhood (the neighborhood is defined as the center of the point, the radius is 0.5mm) around the point to realize the smoothing of the point cloud data, and form a smoother point cloud set with stronger continuity.

[0038] On this basis, for the smoothed point cloud set, the height deviation of each data point and the surrounding data points is further compared. If it is detected that the height value of a point is obviously higher or lower than the overall trend of the neighborhood, the height average value of the field point cloud (the field is defined as the center of the point, the radius is 0.5mm) is smoothed to obtain feature data conforming to the topographic accuracy by locally adjusting the abnormal value.

[0039] Finally, for the above feature data, the point cloud data is converted into a three-dimensional grid structure, thereby improving the consistency and visualization effect of the overall data. In one possible implementation, the grid cell length can be set to 0.015mm, and a triangular grid is constructed based on the spatial position relationship of adjacent points to form a complete curved surface reconstruction model. The structuring process not only facilitates subsequent surface roughness analysis and three-dimensional visualization display, but also provides solid data support for accurate evaluation of aluminum alloy surface quality.

[0040] In step S3, the surface feature refined data is decomposed to extract flatness-related components and roughness-related components, including: The surface feature refined data is analyzed hierarchically to obtain flatness components and roughness components, and a first decomposition data set is obtained by integration; According to the first decomposition data set, the flatness components and the roughness components are processed by feature correlation mapping, and the fluctuation of surface details is detected. If the surface details in the components exceed the preset detail fluctuation threshold range, local adjustment and correction are performed to obtain a second decomposition data set; According to the second decomposition data set, the flatness components and the roughness components are quantitatively processed, and the surface flatness and the surface roughness are extracted. If the surface flatness and the surface roughness do not conform to the preset flatness and roughness threshold values, the corresponding data is deleted to obtain a feature parameter set; The feature parameter set is classified and arranged to obtain the final flatness-related components and roughness-related components.

[0041] Specifically, based on the spatial variation characteristics and the topographic curvature distribution of the data, the original data is stripped into different components. For data processing of the aluminum alloy surface, the surface details are separated layer by layer, the areas with higher flatness are classified as flat components, and the areas with slight undulations or textures are classified as rough components, thereby forming a first decomposed data set. The data set contains two main parts, flatness component and roughness component, which respectively reflect the flatness of the overall aluminum alloy surface and the local microscopic roughness characteristics.

[0042] For the first decomposed data set, the fluctuation of surface details is detected by comparing the distribution of the flat component and the rough component. For example, when the height fluctuation of a certain area in the rough component exceeds the preset threshold (such as 0.2 mm), the system can determine that the area may have surface defects or abnormal signals. At this time, the original topographic data is smoothed and repaired using a neighborhood fitting method to form a second decomposed data set, thereby improving the consistency and accuracy of the overall data. Specifically, a neighborhood window with a radius of 5 pixels centered on each data point is selected, and a Gaussian weighted linear regression algorithm is applied within the window to fit and correct the height data.

[0043] On the basis of the second decomposed data set, the components are quantitatively processed. The surface flatness (such as the regional height standard deviation) and the surface roughness (such as the arithmetic average roughness Ra value) are calculated by a parameter extraction algorithm. Assuming that the preset flatness threshold is 0.1-0.3 mm and the roughness threshold is 1.5-2.5 microns, if a sub-area flatness of 0.4 mm (corresponding to a depression on the aluminum alloy surface) or a roughness Ra=3.2 microns (representing excessive wear) is detected, the data deletion mechanism is triggered. After screening, a feature parameter set is formed, effectively eliminating abnormal data and ensuring that the parameter set meets the industrial quality standards.

[0044] Finally, the feature parameter set is classified and arranged: the flatness-related parameters (such as height deviation gradient, curvature radius) are merged into flatness-related components, and the roughness parameters (such as ripple density, Ra value) are integrated into roughness-related components.

[0045] In step S4, geometric deviation calculation is performed according to the flatness-related components, and if the deviation value exceeds the preset flatness threshold, the area is marked as an abnormal area, and a flatness abnormality distribution map is obtained, including: The flatness-related components are subjected to deviation calculation to obtain a flatness deviation value. If the flatness deviation value exceeds the preset flatness threshold range, the area exceeding the threshold is marked as an abnormal area, and an abnormal area set is obtained. According to the abnormal area set, the abnormal areas are subjected to detail layering processing to obtain layering area data. The layered region data is combined with a preset set of standard parameters and then visualized to generate a flatness anomaly distribution map.

[0046] Specifically, the geometric deviation of the flatness-related components is first calculated. Local deviation quantification is achieved through a mathematical function Ψ, such as the van Albada function Ψ(r) = (r² + r) / (1 + r²). This function calculates based on the height gradient ratio r between adjacent data points, effectively handling regions with abrupt changes in surface gradient. After calculation, a global flatness deviation value matrix is ​​generated. When the deviation value in a certain area exceeds a preset threshold range (e.g., 0.1-0.3 mm according to aviation standards), the system automatically marks that coordinate point as an outlier.

[0047] Subsequently, spatial clustering is performed on the marked outliers, merging adjacent outliers with a spacing less than a preset threshold (e.g., 2mm) into continuous outlier regions, forming an outlier region set. This process uses a depth-first search algorithm to traverse the grid nodes, ensuring that physically adjacent outliers are grouped into the same outlier cell. After the outlier region set is established, detailed layering is further performed using topology decomposition: each outlier region is divided into standard-sized grid cells (e.g., 5×5mm), and quadratic curvature analysis is performed on the sub-cells to calculate the height change rate, generating structured layered region data containing hierarchical deviation data.

[0048] Finally, a graphics mapping engine is used to achieve the visualization output. A deviation value-color coding rule is established: red corresponds to a severe defect area >0.3mm, yellow to a warning area of ​​0.2-0.3mm, and green to a acceptable area <0.2mm. After comparing the layered area data with the tolerance range of the standard parameter library using a matrix, a two-dimensional / three-dimensional thermal distribution map is generated. This distribution map visually displays the spatial location and severity of abnormal areas through color gradients, providing a visual basis for process adjustments.

[0049] In step S5, a local region comparison analysis is performed based on the smoothness anomaly distribution map and the roughness related components to obtain the roughness anomaly correlation results.

[0050] Based on the flatness anomaly distribution map, the abnormal areas are preliminarily divided to obtain the location information of key areas; Extract the roughness-related components corresponding to the key region location information from the roughness-related components, and map them one by one to obtain the target range with large parameter fluctuation amplitude; The target interval is correlated with the trend characteristics to obtain the trend distribution of the roughness parameters; If the trend distribution exceeds the preset roughness threshold range, the trend distribution is fused with the layered region data to obtain a roughness anomaly correlation result.

[0051] Firstly, based on the flatness anomaly distribution map, the marked abnormal area is spatially gridded. Specifically, the abnormal area is divided into several sub-regions according to a preset size (such as 5 square centimeters), and by comparing the flatness deviation value of each sub-region with the average value of the overall distribution, the key positions with a deviation significantly higher than the surrounding area are screened out. For example, when the flatness deviation of a sub-region reaches 0.5 mm, and the average value of the adjacent region is only 0.2 mm, the sub-region is marked as a key region, and its spatial coordinate information is recorded to form a key region position information data set.

[0052] It should be noted that the roughness related component corresponding to the key region position information is extracted from the roughness related component, and is mapped one by one to obtain a target interval with a large parameter fluctuation amplitude. The specific process is to extract the roughness parameter data corresponding to the key region position information from the roughness related component. By spatial mapping technology, the coordinates of the key region are matched with the position points in the roughness component one by one, and the fluctuation amplitude of the roughness parameter in the key region relative to the average value of the adjacent region is calculated. For example, if the roughness parameter of the key region is 3.2 microns, and the average value of the surrounding region is 2.5 microns, the fluctuation amplitude is 0.7 microns; if the fluctuation amplitude exceeds the preset fluctuation threshold (such as 0.3 microns), the region is determined as a target interval with a large parameter fluctuation amplitude, and a target interval distribution map is generated.

[0053] It should be noted that the target interval is associated with the change trend feature to obtain the trend distribution of the roughness parameter. The roughness parameter sequence of the target interval in multiple detections (such as the continuous three detection values are 3.2 microns, 3.5 microns, and 3.8 microns) is obtained, the change direction (such as continuous rise) and slope are analyzed, and the trend distribution data of the roughness parameter is generated. If the trend distribution exceeds the preset roughness threshold range (such as 2.0-3.0 microns), it is determined as an abnormal trend.

[0054] Finally, the above abnormal trend data is fused with the layered information of the flatness abnormal area (such as the layered region data in step S4). For example, if the abnormal area is layered into the surface layer and the subsurface layer in depth, the roughness trend value of the surface layer (3.8 microns) is associated and integrated with the corresponding value of the subsurface layer (3.0 microns) to form a multi-dimensional data set containing spatial position, depth layering and trend feature, i.e. roughness abnormality association result. The result is presented by a visual map, for example, a heat map superimposed with trend arrow marks the change direction of the abnormal area, which provides a basis for subsequent comprehensive quality assessment.

[0055] In step S6, the flatness anomaly distribution map and the roughness abnormality association result are fused to perform comprehensive correlation analysis to obtain a surface quality comprehensive feature matrix.

[0056] According to the flatness anomaly distribution map and the roughness anomaly correlation result, an abnormal region is divided, and distribution trend information in the region is extracted, and the distribution trend information is combined with the correlation of the preset quality index to determine key position data of the abnormal region; According to the key position data, the roughness related component is dynamically mapped and analyzed with the flatness anomaly distribution map to obtain a potential fluctuation interval of surface quality; If the fluctuation interval exceeds a preset quality fluctuation threshold range, the roughness related component and the flatness anomaly distribution map are integrated to generate a comprehensive feature data set; According to the comprehensive feature data set and the distribution trend information, a surface quality comprehensive feature matrix is constructed.

[0057] Specifically, first, based on the flatness anomaly distribution map and the roughness anomaly correlation result, an abnormal region is spatially grid-divided (for example, an aluminum alloy surface is divided by 10 square centimeter units), and distribution trend information in each grid cell is extracted, including flatness deviation change law and roughness fluctuation direction. In combination with the correlation analysis of the preset quality index (such as flatness threshold 0.3 mm and roughness threshold 3.0 μm), the grid deviating from the standard is selected as the key position data. For example, when the flatness deviation of a certain grid reaches 0.6 mm (exceeding the threshold 0.3 mm) and accompanied by an upward trend of roughness, it is marked as a key region.

[0058] Secondly, dynamic mapping analysis is performed on the key position data: the numerical value of the roughness related component is spatially aligned with the flatness anomaly distribution map to establish a parameter correlation model. By calculating the spatial correlation coefficient (such as Pearson correlation coefficient) of the roughness gradient and the flatness deviation, the region with synchronous change of the two is identified, so as to determine the potential fluctuation interval of the surface quality. For example, the roughness value 3.5 μm in the key region is significantly higher than the surrounding average value 2.0 μm, and it is strongly positively correlated (correlation coefficient > 0.8) with the flatness deviation 0.6 mm, so it is determined that the region is a potential fluctuation interval.

[0059] Subsequently, when the parameter variation amplitude of the fluctuation interval exceeds a preset quality fluctuation threshold (such as roughness fluctuation > 0.5 μm), the roughness related component and the flatness anomaly distribution map are integrated: the roughness parameter and the flatness deviation value are superimposed according to the coordinates to generate a three-dimensional vector containing position coordinates, roughness value, and flatness deviation. The vector data is normalized to eliminate dimensional differences; based on vector space clustering (such as K-means algorithm), spatial clustering is performed to form a comprehensive feature data set. For example, the data set after integration of a certain fluctuation interval contains fields such as coordinates (10, 20), roughness 3.8 μm, and flatness deviation 0.7 mm.

[0060] Finally, the surface quality comprehensive feature matrix is constructed in combination with the distribution trend information: the comprehensive feature data set is classified according to the spatial hierarchy (for example, surface layer / subsurface layer). A trend weight is assigned to each hierarchy (for example, surface layer trend weight 0.7, subsurface layer 0.3). A two-dimensional matrix is constructed, with rows representing spatial positions and columns containing roughness, flatness weighted values (calculation formula: feature value = roughness x trend weight + flatness deviation x trend weight) and trend intensity coefficients. The matrix comprehensively represents the surface quality state through quantified multidimensional features (spatial position, hierarchical parameter, trend correlation), providing a structured data basis for subsequent quality evaluation.

[0061] In step S7, the flatness-related component and the roughness-related component are weighted and integrated according to the surface quality comprehensive feature matrix. If the integrated feature value deviates from the preset quality evaluation system standard, it is marked as a quality substandard area, and a quality evaluation distribution map is obtained. An overall surface quality distribution map is obtained. According to the surface quality comprehensive feature matrix, the feature weights of the flatness-related component and the roughness-related component are calculated one by one to obtain a first feature set after weighted integration, and the first feature set is compared with a preset quality standard to obtain a comparison result of the first feature set. If the comparison result of the first feature set exceeds the preset comparison threshold range, the substandard areas in the first feature set are preliminarily marked to obtain a marked area data set. According to the area data set, the substandard areas are associated with the overall distribution map of the surface quality, and abnormal position information in the distribution map is extracted. It is judged whether the abnormal position information conforms to the preset quality evaluation system standard, and a judgment result of the abnormal position information is obtained. According to the judgment result of the abnormal position information, the distribution map is finally adjusted to generate a final quality evaluation distribution map containing substandard area marks.

[0062] Specifically, based on the data of each spatial unit in the surface quality comprehensive feature matrix, the feature weight calculation of the flatness-related component and the roughness-related component is performed. The weight coefficient of the flatness index is set to 0.6, and the weight coefficient of the roughness index is set to 0.4. The comprehensive feature value is calculated by weighted calculation to form a first feature set reflecting the overall quality condition. Taking the detection of an aluminum alloy plate as an example, when the flatness deviation of a unit is 0.7 mm and the roughness Ra value is 3.5 microns, the numerical result of 1.82 is calculated according to the formula "comprehensive feature value = flatness deviation x 0.6 + roughness value x 0.4", which will be used as the quality evaluation benchmark of the unit.

[0063] It should be noted that the first feature set is compared with the preset quality evaluation standard, and the standard sets the upper limit of the flatness deviation threshold value as 0.5 mm, the upper limit of the roughness threshold value as 3.0 microns, and the upper limit of the comprehensive feature value threshold value as 1.5. When the data of the detection unit exceeds any threshold value, it is determined that the quality is not up to standard. The spatial coordinate position, specific flatness value, roughness value and calculated comprehensive feature value of the unit are recorded to form a marked area data set. For example, when the coordinate (15, 20) is detected, because the comprehensive feature value 1.82 exceeds the threshold value 1.5, it is marked as an abnormal unit and the complete parameters are recorded.

[0064] It should be noted that the marked area data set is associated with the overall distribution map of the surface quality in space position, and the corresponding unit in the distribution map is located through coordinate mapping. Cluster analysis is performed on adjacent abnormal units to identify continuous defect areas and divide defect levels according to the exceeding amplitude. In the aluminum alloy plate detection example, when the flatness deviation of the five units around the coordinate (15, 20) is greater than 0.6 mm and the roughness is greater than 3.2 microns, a continuous defect area of 10 square centimeters is formed, which is marked as a first-level serious defect area in the distribution map.

[0065] It should be noted that when generating the final quality evaluation distribution map, color blocks are used to mark abnormal areas on the distribution map base, and different colors represent different defect levels. Key parameter information and process optimization suggestions are added beside the abnormal area to form a visual map that can directly guide production. For example, a red cover layer is displayed in the 10 square centimeter serious defect area, which is marked as "flatness deviation 0.68-0.72 mm, roughness 3.4-3.6 microns", and a specific process adjustment scheme of "precision grinding and polishing treatment: pressure increase by 20%, time increase by 15 seconds" is given.

[0066] In step S8, according to the quality evaluation distribution map, the feature parameters of the quality not up to standard area are extracted, and detection feedback data is generated.

[0067] Obtain the historical adjustment record of the polishing process; According to the quality evaluation distribution map, the feature parameters of the quality not up to standard area are separated and processed, the surface roughness value and the flatness value are extracted, and a preliminary data set is obtained; The preliminary data set is matched with the preset polishing process parameter library, if the surface roughness value and the flatness value in the preliminary data set exceed the preset matching threshold range, the quality not up to standard area is classified and labeled, and the classified feature parameter set is determined; The classified feature parameter set is associated with the adjustment scheme database of the preset polishing process to obtain preliminary feedback data for the quality not up to standard area; According to the preliminary feedback data, it is fused with the historical adjustment record of the polishing process to generate final detection feedback data.

[0068] Firstly, the historical polishing process adjustment record is obtained, which contains parameter optimization scheme and actual improvement effect data of different aluminum alloy parts under similar surface defect conditions. For example, in a certain automobile aluminum alloy hub production line, the historical record shows that when the local roughness value reaches 3.5 microns, after adopting the adjustment scheme of "polishing pressure is increased by 0.05 MPa and processing time is extended by 20 seconds", the roughness can be reduced to below 3.0 microns.

[0069] Based on the final quality evaluation distribution map, the feature parameters of the quality unqualified area marked in the map are extracted. According to the coordinate positioning of the abnormal area, the surface roughness measured value and the flatness deviation measured value of the area are separated to form a preliminary data set containing three parameters of spatial position, roughness and flatness. Taking the detection of a certain aluminum alloy engine cover part as an example, the 100 square centimeter area with coordinates X10 to X20 and Y30 to Y40 in the distribution map is marked as a red warning area, and the average roughness in this area is extracted as 3.8 microns and the maximum flatness deviation is 0.75 millimeters, and the data set record is "area coordinates: X10-20 / Y30-40; roughness: 3.8 microns; flatness: 0.75 millimeters".

[0070] The preliminary data set is matched with the preset polishing process parameter library, and the parameter library sets the threshold standard as the upper limit of roughness 3.0 microns and the upper limit of flatness deviation 0.5 millimeters. When the measured value exceeds the threshold, the defect classification is performed according to the exceeding standard: the roughness measured value in the range of 3.0-3.6 microns or the flatness deviation in the range of 0.5-0.6 millimeters is defined as mild defect; roughness 3.6-4.5 microns or flatness 0.6-0.75 millimeters is defined as moderate defect; any parameter exceeding the moderate range is defined as severe defect. The aforementioned engine cover part is classified and marked as severe defect due to roughness 3.8 microns (exceeding standard 26.7%) and flatness 0.75 millimeters (exceeding standard 50%), and the feature parameter set record is "defect level: severe; roughness: 3.8 microns; flatness: 0.75 millimeters".

[0071] The classified feature parameter set is associated with a polishing process adjustment scheme database, and a basic adjustment scheme is matched according to the defect level. For severe defects, a 'fine grinding treatment + parameter reconstruction' framework is used, and the parameters are fine-tuned according to the specific over-standard value: for each 0.1 microns of roughness over-standard, 5% polishing pressure is increased, and for each 0.1 millimeters of flatness over-standard, 10 seconds of processing time is extended. Continuing the hood case, the basic scheme is 'fine grinding, pressure increase by 30%, time extension by 25 seconds', and according to the roughness over-standard value of 0.8 microns, an additional 40% pressure adjustment (0.8 / 0.1 x 5%) is made, generating preliminary feedback data 'fine grinding treatment; pressure adjustment: +70% (30%+40%); time adjustment: +25 seconds'.

[0072] The preliminary feedback data is optimized by fusing historical adjustment records, and successful adjustment cases of similar parts under the same defect level are retrieved. If the historical records show that the 'pressure +70%' scheme used for the same type of hood has caused surface burning, and the optimized 'two-time polishing (first time pressure +50% time +15 seconds, second time pressure +30% time +10 seconds)' scheme effectively solves the problem, then the feedback data is corrected as 'final scheme: fine grinding is performed twice, first time pressure +50% time +15 seconds, second time pressure +30% time +10 seconds; expected improvement: roughness ≤3.0 μm, flatness ≤0.5 mm'.

[0073] In a second aspect, as shown in Figure 2 The application provides a quality detection system for aluminum alloy polishing, comprising: A data acquisition module for acquiring a preliminary surface feature data set; A data denoising module for denoising the preliminary surface feature data set to obtain refined surface feature data; A feature decomposition module for hierarchical analysis of the refined surface feature data to extract flatness-related components and roughness-related components; A flatness anomaly detection module for calculating geometric deviation based on the flatness-related components, and marking abnormal areas if the deviation value exceeds the preset flatness threshold range to obtain a flatness anomaly distribution map; A roughness anomaly analysis module for local area comparison analysis based on the flatness anomaly distribution map and the roughness-related components to obtain roughness anomaly correlation results; A feature fusion module for comprehensive correlation analysis of two types of data based on the flatness anomaly distribution map and the roughness anomaly correlation results to obtain a surface quality comprehensive feature matrix; The quality evaluation module is configured to integrate the flatness and roughness features by weighting according to the surface quality comprehensive feature matrix, and mark the quality non-compliance region if the integrated feature value deviates from a preset quality evaluation system standard to obtain a quality evaluation distribution map. The feedback generation module is configured to extract feature parameters of the quality non-compliance region according to the quality evaluation distribution map, and generate detection feedback data.

[0074] In a third aspect, the present application further provides an electronic device, comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the quality detection method of the aluminum alloy polishing according to any one of the above.

[0075] In a fourth aspect, the present application further provides a computer readable storage medium, comprising a stored computer program, wherein the computer program controls a device where the computer readable storage medium is located to execute the quality detection method of the aluminum alloy polishing according to any one of the above when the computer program is running.

Claims

1. A method for quality detection of aluminum alloy polishing, characterized by, The method comprises the following steps: obtaining a preliminary surface feature data set; denoising the preliminary surface feature data set to obtain surface feature refined data; performing hierarchical analysis on the surface feature refined data to extract flatness-related components and roughness-related components; performing geometric deviation calculation according to the flatness-related components, and if the deviation value exceeds the preset flatness threshold, marking it as an abnormal area to obtain a flatness abnormality distribution map; performing local area comparison analysis according to the flatness abnormality distribution map and the roughness-related components to obtain a roughness abnormality correlation result; integrating the flatness abnormality distribution map and the roughness abnormality correlation result to perform comprehensive correlation analysis and obtain a surface quality comprehensive feature matrix; performing weighted integration on the flatness-related components and the roughness-related components according to the surface quality comprehensive feature matrix, and if the integrated feature value deviates from the preset quality evaluation system standard, marking it as a quality substandard area to obtain a quality evaluation distribution map; extracting the feature parameters of the quality substandard area according to the quality evaluation distribution map to generate detection feedback data.

2. The method for quality detection of aluminum alloy polishing according to claim 1, characterized in that, The method comprises the following steps: obtaining a preliminary surface feature data set; obtaining three-dimensional topography data of the surface of the aluminum alloy, i.e., first topography data; performing denoising and filtering on the first topography data to obtain second topography data; performing grid processing on the second topography data to obtain third topography data; 3. The method for quality detection of aluminum alloy polishing according to claim 1, characterized in that, performing surface roughness quantification analysis and feature extraction on the third topography data to obtain a preliminary surface feature data set. The method comprises the following steps: performing hierarchical separation on the surface noise in the preliminary surface feature data set to extract less disturbed point cloud information to obtain optimized point cloud data; if there is data in the optimized point cloud data that exceeds the preset fluctuation threshold range, performing interpolation correction on the data to obtain a smoothed point cloud set; performing abnormality detection on the smoothed point cloud set, and if it is found that the data points have large deviation from the overall surface details, performing local optimization to obtain feature data that meets the topography accuracy; 4. The method for quality detection of aluminum alloy polishing according to claim 1, characterized in that, performing structured processing on the feature data to convert the discrete point cloud data into a structured grid form to obtain optimized surface feature refined data. The method comprises the following steps: performing hierarchical analysis on the surface feature refined data to obtain flat components and rough components, and integrating to obtain a first decomposition data set; performing feature correlation mapping processing on the flat components and the rough components according to the first decomposition data set to detect the fluctuation of surface details, and if it is found that the surface details in the components exceed the preset detail fluctuation threshold range, performing local adjustment and correction to obtain a second decomposition data set; performing quantification processing on the flat components and the rough components according to the second decomposition data set, and extracting surface flatness and surface roughness, and if the surface flatness and the surface roughness do not meet the preset flatness and roughness threshold, deleting the corresponding data to obtain a feature parameter set; The feature parameter set is classified and arranged to obtain a final flatness related component and a roughness related component.

5. The method for quality detection of aluminum alloy polishing according to claim 1, characterized in that, According to the flatness related component, a geometric deviation is calculated, and if the deviation value exceeds a preset flatness threshold, the area is marked as an abnormal area to obtain a flatness abnormal distribution map, including: A deviation of the flatness related component is calculated to obtain a flatness deviation value. If the flatness deviation value exceeds a preset flatness threshold range, the area exceeding the threshold is marked as an abnormal area to obtain an abnormal area set. According to the abnormal area set, the abnormal area is processed in detail to obtain layered area data. The layered area data is combined with a preset standard parameter set for visual processing to generate a flatness abnormal distribution map.

6. The method for quality detection of aluminum alloy polishing according to claim 5, characterized in that, According to the flatness abnormal distribution map and the roughness related component, a local area comparison analysis is performed to obtain a roughness abnormal correlation result, including: According to the flatness abnormal distribution map, the abnormal area is preliminarily divided to obtain key area position information. From the roughness related component, the roughness related component corresponding to the key area position information is extracted and mapped one by one to obtain a target interval with a larger parameter fluctuation amplitude. The target interval is associated with a change trend feature to obtain a trend distribution of the roughness parameter. If the trend distribution exceeds a preset roughness threshold range, the trend distribution is fused with the layered area data to obtain a roughness abnormal correlation result.

7. The method for quality detection of aluminum alloy polishing according to claim 1, wherein, The flatness abnormal distribution map and the roughness abnormal correlation result are fused for comprehensive correlation analysis to obtain a surface quality comprehensive feature matrix, including: According to the flatness abnormal distribution map and the roughness abnormal correlation result, the abnormal area is divided, and the distribution trend information in the area is extracted, the correlation between the distribution trend information and a preset quality index is determined to determine the key position data of the abnormal area. According to the key position data, the roughness related component and the flatness abnormal distribution map are dynamically mapped and analyzed to obtain a potential fluctuation interval of the surface quality. If the fluctuation interval exceeds a preset quality fluctuation threshold range, the roughness related component and the flatness abnormal distribution map are integrated to generate a comprehensive feature data set. According to the comprehensive feature data set and the distribution trend information, a surface quality comprehensive feature matrix is constructed.

8. The method for quality detection of aluminum alloy polishing according to claim 1, characterized in that, According to the surface quality comprehensive feature matrix, the features of flatness and roughness are weighted and integrated, and if the integrated feature value deviates from a preset quality evaluation system standard, the area is marked as a quality substandard area to obtain a quality evaluation distribution map, including: An overall surface quality distribution map is obtained. According to the surface quality comprehensive feature matrix, the feature weights of the flatness related component and the roughness related component are calculated one by one to obtain a first feature set after weighted integration, and the first feature set is compared with a preset quality standard to obtain a comparison result of the first feature set. If the comparison result of the first feature set exceeds the preset comparison threshold range, the substandard area in the first feature set is preliminarily marked to obtain a marked area data set; According to the area data set, the substandard area is associated with the overall distribution map of the surface quality for processing, abnormal position information in the distribution map is extracted, whether the abnormal position information meets the preset quality evaluation system standard is judged, and a judgment result of the abnormal position information is obtained; According to the judgment result of the abnormal position information, the distribution map is finally adjusted to generate a final quality evaluation distribution map containing substandard area marking.

9. The method for quality detection of aluminum alloy polishing according to claim 1, wherein, According to the quality evaluation distribution map, the feature parameters of the quality substandard area are extracted to generate detection feedback data, including: Obtaining the historical adjustment record of the polishing process; According to the quality evaluation distribution map, the feature parameters of the quality substandard area are separated for processing to extract the surface roughness value and flatness value to obtain a preliminary data set; If the surface roughness value and flatness value in the preliminary data set exceed the preset matching threshold range, the quality substandard area is classified and marked to determine a classified feature parameter set; The classified feature parameter set is associated with the adjustment scheme database of the preset polishing process to obtain preliminary feedback data for the quality substandard area; According to the preliminary feedback data, the historical adjustment record of the polishing process is fused to generate final detection feedback data.

10. A quality detection system for aluminum alloy polishing, characterized by, Including: A data acquisition module is configured to acquire a preliminary surface feature data set; A data denoising module is configured to perform denoising processing on the preliminary surface feature data set to obtain surface feature refined data; A feature decomposition module is configured to perform hierarchical analysis on the surface feature refined data to extract flatness-related components and roughness-related components; A flatness anomaly detection module is configured to perform geometric deviation calculation according to the flatness-related components, and if the deviation value exceeds the preset flatness threshold range, the area is marked as an abnormal area to obtain a flatness anomaly distribution map; A roughness anomaly analysis module is configured to perform local area comparison analysis according to the flatness anomaly distribution map and the roughness-related components to obtain a roughness anomaly correlation result; A feature fusion module is configured to fuse the flatness anomaly distribution map and the roughness anomaly correlation result for comprehensive correlation analysis to obtain a surface quality comprehensive feature matrix; A quality evaluation module is configured to weight and integrate the flatness and roughness features according to the surface quality comprehensive feature matrix, and if the integrated feature value deviates from the preset quality evaluation system standard, the area is marked as a quality substandard area to obtain a quality evaluation distribution map; A feedback generation module is configured to extract the feature parameters of the quality substandard area according to the quality evaluation distribution map to generate detection feedback data.

Images