Metal washer full-automatic detection method based on industrial vision

Abstract

The invention relates to the technical field of metal washer detection, and discloses a metal washer full-automatic detection method based on industrial vision, which comprises the following steps: detecting a target washer to preliminarily judge whether the target washer is qualified or not, forming a preliminary detection conclusion, performing secondary analysis or preliminary misjudgment analysis according to the preliminary detection conclusion, and determining whether the target washer is qualified or not. According to the full-automatic detection method for the metal washer based on the industrial vision, blindness of traditional manual sampling inspection is replaced, quality risks are converted from'post discovery 'to'pre-prevention', the misjudgment rate is reduced, waste of qualified products is reduced, the quality cost of enterprises is reduced, the credibility of a detection system is improved, and the detection efficiency is improved. The method has the advantages that the internal defect detectability is improved, the problem of traditional visual leak detection is quickly screened out, the early failure after the metal gasket is installed is avoided, the detection intensity is automatically matched, the detection cost is reduced, the one-step excessive detection is avoided, the optimal configuration of detection resources is realized, and the precision of a detection system is promoted to be continuously improved.

Description

Technical Field

[0001] This invention relates to the field of metal gasket inspection technology, specifically to a fully automated metal gasket inspection method based on industrial vision. Background Technology

[0002] Metal gaskets, as a fundamental and widely used sealing and fastening component, are extensively applied in industries such as automotive, aerospace, machinery manufacturing, and electronics. Their quality (including dimensional accuracy, shape defects, and surface flaws) directly affects the sealing performance, stability, and safety of assemblies. Traditional inspection methods mainly rely on manual sampling using tools such as calipers, plug gauges, and micrometers, which suffers from problems such as low efficiency, high cost, strong subjectivity, inconsistent standards, difficulty in detecting minute defects, lack of data traceability, and potential damage to products due to contact measurement. The fully automated metal gasket inspection method based on industrial vision is realized through a modular and integrated industrial vision system. The entire inspection process is completed in a fully automated production line without human intervention. It aims to provide a highly efficient, accurate, stable, and non-contact fully automated metal gasket inspection method to solve the problems of low efficiency, insufficient accuracy, and inconsistent standards of traditional manual inspection. It achieves 100% inspection of the size, shape, and surface quality of metal gaskets and establishes a complete quality data archive.

[0003] Existing fully automated inspection methods for metal washers based on industrial vision cannot infer whether there are potential defects inside the washer from defects on the washer surface, nor can they infer whether there are potential defects inside the washer by sampling inspected washers. They also cannot promptly determine whether the inspection conclusions based on the presence of defects on the washer surface are misjudgments. This can easily lead to over-inspection with a "one-size-fits-all" approach, resulting in high inspection costs, low inspection efficiency, and certain limitations in practicality. Summary of the Invention

[0004] This invention provides a fully automated inspection method for metal washers based on industrial vision, which helps to solve the problems mentioned in the background art.

[0005] This invention provides the following technical solution: a fully automated inspection method for metal washers based on industrial vision, comprising:

[0006] Obtain the target washer;

[0007] By inspecting the target gasket using industrial vision, a preliminary judgment is made as to whether the target gasket is qualified, and a preliminary inspection conclusion is formed.

[0008] If the target gasket is initially determined to be qualified, a second analysis will be conducted.

[0009] If the target gasket is initially determined to be unqualified, a preliminary misjudgment analysis should be conducted on the preliminary test results.

[0010] If the initial assessment indicates a misjudgment, a second analysis will be conducted.

[0011] If the initial assessment determines that there is no misjudgment, then a defect analysis will be conducted.

[0012] As an optional solution to the fully automated metal gasket inspection method based on industrial vision described in this invention, the method involves: inspecting the target gasket using industrial vision to preliminarily determine whether the target gasket is qualified, specifically:

[0013] Acquire an image of the target washer and designate it as the target image;

[0014] Preprocess the target image;

[0015] Geometrically locate the target image;

[0016] Extract defect features;

[0017] Calculate the overall defect score for each candidate region;

[0018] By region determination function For each candidate region, determine whether it has a defect;

[0019] like If so, the candidate region is determined to be defective;

[0020] like If so, the candidate region is determined to be free of defects;

[0021] Preliminary defect judgment function A preliminary assessment is made to determine whether the target gasket is qualified.

[0022] like If so, the target washer is preliminarily determined to be qualified;

[0023] like If so, it is preliminarily determined that the target washer is unqualified.

[0024] As an optional solution to the fully automated metal gasket inspection method based on industrial vision described in this invention, the secondary analysis is specifically performed as follows:

[0025] Perform anomaly analysis on the target gasket and generate anomaly judgment data;

[0026] Obtain sampling data;

[0027] Calculate the multi-parameter comprehensive score of the target washer;

[0028] Through the exception detection function Determine if there are any abnormalities inside the target gasket;

[0029] like If so, it is determined that there is an abnormality inside the target washer;

[0030] like If so, it is determined that there is no abnormality inside the target washer;

[0031] By comprehensive judgment function A comprehensive judgment is made as to whether the target gasket is qualified;

[0032] like If so, the target washer is ultimately deemed qualified;

[0033] like If so, then defect analysis will be performed.

[0034] As an optional solution to the fully automated metal gasket detection method based on industrial vision described in this invention, the method involves: performing anomaly analysis on the target gasket and generating anomaly judgment data, specifically:

[0035] Collect the weight of the target washer;

[0036] Calculate the relative deviation value;

[0037] Weight determination function To determine if the weight of the target washer is abnormal;

[0038] like If so, the weight of the target washer is determined to be abnormal;

[0039] like If the weight of the target washer is normal, then it is determined that the weight of the target washer is normal.

[0040] Collect the center of gravity position of the target washer;

[0041] Calculate the center of gravity offset;

[0042] Using the centroid determination function To determine whether the center of gravity of the target washer is abnormal;

[0043] like If so, the center of gravity of the target washer is determined to be abnormal;

[0044] like If the target washer's center of gravity is normal, then it is determined that the target washer's center of gravity is normal.

[0045] Collect the compressive stiffness of the target washer;

[0046] Stiffness determination function To determine whether the stiffness of the target washer is abnormal;

[0047] like If so, the stiffness of the target washer is determined to be abnormal;

[0048] like If the stiffness of the target washer is normal, then it is determined that the stiffness of the target washer is normal.

[0049] Collect the natural frequency of the target washer;

[0050] Frequency determination function To determine whether the natural frequency of the target washer is abnormal;

[0051] like If so, the natural frequency of the target washer is determined to be abnormal;

[0052] like If the natural frequency of the target washer is normal, then it is determined that the natural frequency of the target washer is normal.

[0053] Collect the damping ratio of the target washer;

[0054] Damping criterion function To determine whether the damping ratio of the target washer is abnormal;

[0055] like If so, the damping ratio of the target washer is determined to be abnormal;

[0056] like If the damping ratio of the target washer is normal, then it is determined that the damping ratio is normal.

[0057] As an optional solution to the fully automated metal gasket inspection method based on industrial vision described in this invention, the acquisition of sampling data specifically includes:

[0058] Set the sampling size;

[0059] Perform stratified sampling;

[0060] Surface defects were verified on the sampled metal gaskets;

[0061] Internal defect detection was performed on the sampled metal gaskets;

[0062] Data analysis and batch quality assessment were performed on the sampled metal gaskets.

[0063] Based on the sampling comprehensive judgment function To determine whether the sampled gaskets meet the normal requirements;

[0064] like If so, the sampled gaskets are deemed to be in normal condition;

[0065] like If so, it is determined that the sampled gaskets do not meet the normal requirements.

[0066] As an optional solution to the fully automated metal gasket detection method based on industrial vision described in this invention, the preliminary detection conclusions are analyzed for potential errors, specifically as follows:

[0067] Washers that are deemed abnormal are classified as abnormal washers.

[0068] The faulty gasket was placed back into the same testing equipment for repeated testing;

[0069] The defective gaskets were re-inspected using a separate testing system.

[0070] The threshold check function is used to check whether the threshold setting is abnormal.

[0071] The algorithm and model validation function is used to check for anomalies in the feature extraction algorithm and the illumination compensation algorithm.

[0072] The vibration interference detection function is used to check whether there are any abnormalities in the workshop vibration.

[0073] The light source stability check function is used to check for any abnormalities in illuminance fluctuations.

[0074] The electromagnetic interference verification function is used to check for the presence of electromagnetic interference.

[0075] Analyze parameters for incorrect judgments.

[0076] As an optional solution to the fully automated metal gasket detection method based on industrial vision described in this invention, the following is included: performing misjudgment parameter analysis, specifically:

[0077] Plot the time / location distribution of defects within the batch;

[0078] Analyze the defect distribution of abnormal washers;

[0079] Verify the function using OC curves. To determine if there are any abnormalities in the detection system;

[0080] like If so, it is determined that the detection system is abnormal;

[0081] like If so, the detection system is considered to be functioning normally;

[0082] To trace production records and determine whether the records are complete and whether the production process is stable;

[0083] Failure Mode Analysis Function To determine whether there is a misjudgment of production actions;

[0084] Calculate the preliminary misjudgment comprehensive judgment result;

[0085] Based on the preliminary comprehensive judgment function To determine whether there was a misjudgment in the initial assessment;

[0086] like If so, it is preliminarily determined that there is a misjudgment;

[0087] like Therefore, it can be preliminarily determined that there is no misjudgment.

[0088] As an optional solution to the fully automated metal gasket inspection method based on industrial vision described in this invention, defect analysis is performed, specifically:

[0089] Analyze the morphology of surface defects based on the visual inspection results;

[0090] Perform three-dimensional depth information analysis on surface defects;

[0091] Abnormal gaskets are detected by ultrasonic waves to determine whether the ultrasonic data of the abnormal gaskets is abnormal.

[0092] The abnormal gasket is inspected using X-rays to determine whether the X-ray data of the abnormal gasket is abnormal.

[0093] By performing magnetic particle testing on abnormal washers, it can be determined whether the magnetic particle data of the abnormal washers is abnormal.

[0094] Comprehensive assessment of execution deficiencies.

[0095] As an optional solution to the fully automated metal gasket inspection method based on industrial vision described in this invention, the following is a method for performing a comprehensive defect assessment:

[0096] Assess the impact of defects on service life;

[0097] Assess the impact of the defect on normal use;

[0098] Comprehensive conclusions regarding computational defects;

[0099] Defect comprehensive judgment function To determine whether the defective gasket is ultimately qualified;

[0100] like If so, the abnormal gasket is deemed to be ultimately qualified;

[0101] like If the gasket is found to be defective, it will be deemed unqualified.

[0102] The present invention has the following beneficial effects:

[0103] 1. This fully automated metal gasket inspection method based on industrial vision performs image acquisition, preprocessing, ROI extraction, defect feature extraction and scoring on the gasket to determine whether there are defects on the surface of the metal gasket. If no defects are found, the target gasket is initially determined to be unqualified; if defects are found, the target gasket is initially determined to be qualified, and then secondary analysis is performed. Macroscopic parameters such as weight, center of gravity, stiffness, and sound frequency of the target gasket are measured to determine whether these parameters are normal. Furthermore, the analysis of sampled gaskets is obtained; that is, the inspected gaskets are sampled and carefully inspected to determine whether the sampled gaskets are normal. If the macroscopic parameters of the target gasket are all normal and the sampled gaskets are also normal, the target gasket is ultimately deemed qualified. If any parameter is abnormal, defect analysis is performed. Based on Otsu adaptive thresholding and connected component analysis, the detection speed is improved, the false detection rate is reduced, and high-speed online full inspection is achieved. This replaces the inefficiency and instability of manual visual inspection, providing a "second line of defense" for visual inspection, discovering subtle internal defects, avoiding the risk of missed detection, improving the confidence of the judgment conclusion of the target gasket after reference sampling analysis, significantly reducing the cost of destructive testing, and ensuring the scientific nature of batch quality judgment.

[0104] 2. This fully automated metal gasket inspection method based on industrial vision performs a false positive analysis if the target gasket is initially determined to be unqualified. This analysis involves repeated inspections, cross-validation with multiple devices, threshold checks, environmental interference investigation, and batch statistical analysis to determine if the initial determination of unqualified gasket is a false positive. If a false positive is found, a secondary analysis is performed; otherwise, a defect analysis is conducted. This three-step method of "re-inspection verification + parameter verification + environmental investigation" reduces the false positive rate, quickly identifies and eliminates systemic interference sources such as vibration, lighting, and inappropriate thresholds, ensuring inspection stability.

[0105] 3. This fully automated metal gasket inspection method based on industrial vision can perform defect analysis if the macroscopic parameters of the target gasket are abnormal, or if the sampled gasket is abnormal, or if the initial determination that the target gasket is unqualified is not a misjudgment. This involves 3D depth measurement of the defective area, and inspection of the gasket's interior using ultrasound / X-ray / CT. By combining surface defect characteristics with internal inspection results, the impact on normal use and service life is assessed. If the gasket's use or service life is affected, it is deemed unqualified and defect details are output. If it is unaffected, it is deemed qualified. Combining 3D depth measurement with ultrasound / X-ray / CT identifies internal defects in the gasket, improves inspection accuracy, provides definitive conclusions, and clarifies the nature of the defect (surface / internal / connected), providing a basis for subsequent handling (scrapping / rework). Attached Figure Description

[0106] Figure 1 This is a flowchart of the fully automated metal gasket detection method based on industrial vision according to the present invention. Detailed Implementation

[0107] 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.

[0108] Example 1: A fully automated inspection method for metal washers based on industrial vision (see [reference]). Figure 1 ,include:

[0109] Obtain the target washer;

[0110] By inspecting the target gasket using industrial vision, a preliminary judgment is made as to whether the target gasket is qualified, i.e. whether there are any appearance defects, thus forming a preliminary inspection conclusion.

[0111] If the target gasket is initially determined to be qualified, a second analysis will be conducted.

[0112] If the target gasket is initially determined to be unqualified, a preliminary misjudgment analysis should be conducted on the preliminary test results.

[0113] If the initial assessment indicates a misjudgment, a second analysis will be conducted.

[0114] If the initial assessment determines that there is no misjudgment, then a defect analysis will be conducted.

[0115] Specifically, industrial vision is used to inspect the target gaskets and make a preliminary judgment on whether the target gaskets are qualified.

[0116] Acquire an image of the target washer and designate it as the target image;

[0117] Preprocess the target image:

[0118] The target image is subjected to Gaussian filtering using a 5×5 kernel to remove noise, resulting in a denoised image. : ;in, This is the convolution operator, indicating that a Gaussian kernel is used. In the image Slide the slider upwards and perform a weighted summation to achieve smoothing filtering. Represents the target image. The Gaussian kernel function (two-dimensional Gaussian distribution) is used to generate smoothing filters. The larger the image, the blurrier it becomes. The specific formula is: ,in, The standard deviation of the Gaussian kernel controls the smoothing degree, such as... A value of 1.5 indicates that the center of the core has a high weight and the edge has a low weight; a 5×5 core covers approximately 3... scope, Represents the natural exponential function This is used to generate a Gaussian-distributed bell curve, so that pixels farther from the center have smaller weights. and These are pixel coordinates, representing the horizontal and vertical offsets from the center of the convolution kernel;

[0119] The denoised image is converted to grayscale according to the ITU-R BT.601 standard to obtain the grayscale image. :

[0120] ;in, , , These are the red, green, and blue color channel values ​​(0-255), representing the three components of each pixel in a color image. , , According to the ITU-R BT.601 standard weights, the contribution ratio of each channel when converting RGB to grayscale is: the human eye is most sensitive to green light, followed by red light, and least sensitive to blue light;

[0121] The optimal threshold is automatically calculated using the Otsu algorithm. Convert grayscale image to generate binary image : ;in, Represents the binary image in coordinates Pixel value at that location, Represents the grayscale image in coordinates Pixel value at that location, The optimal threshold is automatically calculated using the Otsu algorithm. Used to generate binary images. and These are the binarized pixel values, where 255 represents white (foreground / defect) and 0 represents black (background), facilitating subsequent connected component analysis.

[0122] Geometric localization of the target image:

[0123] Extract the gasket profile using the Canny operator;

[0124] The center point is determined by fitting the inner and outer circles using the least squares method. and radius , ;

[0125] Constructing and generating a ring mask: ; ;in, This is a mask function, with a value of 1 within the ROI region and 0 outside the region, used to mask irrelevant areas. For indicator functions (Boolean functions), when When the internal condition is true, it is satisfied. Returns 1 if the condition is met, otherwise returns 0. The mathematical definition of a circular ROI, i.e., pixels. To the center The distance is between the radii of the inner and outer circles. For the ROI region image, only the circular area is retained, and the remaining pixels are set to 0. For the mask matrix, and For the same size, elements are either 0 or 1. Implement masking operations using the Hadamard product (element-wise multiplication). This indicates that the pixels outside the mask will be cleared to zero.

[0126] Extracting defect features:

[0127] Mark all connected regions within the binarized ROI region: ;in, The connected component labeling result is a matrix of the same size as the input image, where each pixel value is the number of its connected component (0 represents the background). This is a connected component labeling algorithm that groups adjacent foreground pixels (with a value of 255) into the same region and returns a labeling matrix.

[0128] Calculate the area of ​​the connected region, denoted as . : ;in, For the first A set of pixels in connected components, containing all pixels labeled as pixel coordinates, Represents pixel coordinates Belongs to the Connected components, Represents statistical connected components The total number of pixels within, i.e., the area ;

[0129] Calculate the contrast within a connected region, denoted as : ;in, The grayscale mean of the background area is used to calculate the contrast difference between the defect and the background.

[0130] Calculate the roundness, denoted as The circularity refers to how closely the shape of a region approximates a circle; the closer the value is to 1, the more rounded it is, and the smaller the value, the more irregular the shape. ;in, For the first The square of the perimeter of each connected region is used for circularity calculation. This represents the calculation of circularity, where the area of ​​a circle is... Circumference is When perfectly round When the shape is not circular ;

[0131] Calculate the comprehensive defect score for each candidate region, denoted as . The candidate regions are each foreground region obtained after connected component analysis, i.e., potential defect regions: ;in, , , These are weighting coefficients, which control the importance of area, contrast, and shape in defect scoring, respectively, and their sum is 1. , , The values ​​are 0.4, 0.3, and 0.3 respectively. This is the maximum permissible defect area, used to normalize the area; defects exceeding this value receive a score of 1. This is the grayscale value normalization coefficient, used to adjust the contrast. (0-255) is converted to the range 0-1;

[0132] By region determination function For each candidate region, determine whether it has defects: ;in, This is the defect determination threshold, used to determine whether a defect exists in the candidate region;

[0133] like If so, the candidate region is determined to be defective;

[0134] like If so, the candidate region is determined to be free of defects;

[0135] Preliminary defect judgment function A preliminary assessment is made to determine whether the target washer is qualified. ;in, This represents the total number of connected components, i.e., the number of candidate regions. To find the sum index, traverse from 1 to... , This represents the total number of regions identified as defects across all candidate regions, i.e., the total number of defects.

[0136] like If so, the target washer is preliminarily determined to be qualified;

[0137] like If so, it is preliminarily determined that the target washer is unqualified.

[0138] Example 2 is an improvement on Example 1. This example involves a secondary analysis of the fully automated metal gasket detection method based on industrial vision.

[0139] Perform anomaly analysis on the target gasket and generate anomaly judgment data;

[0140] Obtain sampling data;

[0141] Calculate the multi-parameter comprehensive score of the target washer, denoted as : ;

[0142] Through the exception detection function Determine if there are any abnormalities inside the target washer:

[0143] ;

[0144] like If so, it is determined that there is an abnormality inside the target washer;

[0145] like If so, it is determined that there is no abnormality inside the target washer;

[0146] By comprehensive judgment function A comprehensive judgment is made as to whether the target washer is qualified:

[0147] ;

[0148] like If so, the target washer is ultimately deemed qualified;

[0149] like If so, then defect analysis will be performed.

[0150] Specifically, anomaly analysis is performed on the target gasket to generate anomaly judgment data, as follows:

[0151] The weight of the target washer is collected and recorded as follows. ;

[0152] Calculate the relative deviation value, denoted as : ;in, The nominal mass, i.e. the design value, is used to calculate the relative deviation value;

[0153] Weight determination function Determine if the weight of the target washer is abnormal: ;in, The relative quality deviation threshold is the upper limit of the allowable deviation between the measured quality and the nominal quality. When defects such as internal holes and looseness cause the weight loss of the gasket to exceed 0.5%, the weight is judged to be abnormal and further internal inspection is triggered.

[0154] like If so, the weight of the target washer is determined to be abnormal;

[0155] like If the weight of the target washer is normal, then it is determined that the weight of the target washer is normal.

[0156] Collect the center of gravity position of the target washer: , ;in, These are the coordinates of the target washer's center of gravity, i.e., the coordinates of the target washer's center of gravity position. For the first The supporting force at each support point is the force value measured by three sensors in the three-point weighing method. According to the principle of torque balance, the greater the force, the closer the center of gravity is to that support point. This force is used to calculate the coordinates of the center of gravity. The support points are the positions of three weighing sensors arranged in an equilateral triangle, which provide the reference positions for torque calculation. The center of gravity is deduced by measuring the force at each point. The coordinates of the support points;

[0157] Calculate the center of gravity offset, denoted as : ;in, The coordinates of the gasket's geometric center are used as a reference point to calculate the offset of the center of gravity relative to the geometric center. To determine whether the mass distribution is uniform;

[0158] Using the centroid determination function Determine if the center of gravity of the target washer is abnormal: ;in, This is the center of gravity offset threshold, used to determine whether the center of gravity of the target washer is abnormal. For example, for a washer with a diameter of 50mm, if... Then it is abnormal;

[0159] like If so, the center of gravity of the target washer is determined to be abnormal;

[0160] like If the target washer's center of gravity is normal, then it is determined that the target washer's center of gravity is normal.

[0161] Collect the compressive stiffness of the target washer:

[0162] Measure the force-displacement curve using a universal testing machine: ;in, Compressive stiffness is the material's ability to resist deformation. The applied compressive force, i.e., the load applied by the universal testing machine. For compressive displacement, that is, under force The amount of deformation of the washer under action, and the slope of the force-displacement curve, is the stiffness.

[0163] Stiffness determination function Determine if the stiffness of the target washer is abnormal: ;in, The design standard stiffness value (the theoretical stiffness without defects) is calculated from material parameters and geometric dimensions and used as a benchmark for comparing with measured stiffness. Compared with the design standard stiffness value A drop exceeding 15% indicates a serious internal defect.

[0164] like If so, the stiffness of the target washer is determined to be abnormal;

[0165] like If the stiffness of the target washer is normal, then it is determined that the stiffness of the target washer is normal.

[0166] The natural frequency of the target washer is collected and denoted as . : ;in, This indicates finding the frequency corresponding to the maximum value of a function, i.e., finding the frequency coordinates of the point with the maximum amplitude in the frequency spectrum, which is used to determine the natural vibration frequency of the washer. Indicates the sound signal Perform a Fast Fourier Transform to convert the time-domain signal into a frequency-domain spectrum, which can be used to analyze the frequency components of the signal and find the resonance peaks.

[0167] Frequency determination function Determine if the natural frequency of the target washer is abnormal: ;in, The design standard natural frequency (the theoretical frequency under defect-free conditions) is used as a benchmark for comparison with measured frequencies. With the inherent frequency of the design standard , The fundamental frequency deviation threshold is set at 3%, indicating that internal defects have caused changes in stiffness or mass.

[0168] like If so, the natural frequency of the target washer is determined to be abnormal;

[0169] like If the natural frequency of the target washer is normal, then it is determined that the natural frequency of the target washer is normal.

[0170] The damping ratio of the target washer was collected by free decaying vibration, and denoted as . : ;in, For the first The amplitude (peak value) of each vibration cycle. For the first The amplitude of each vibration cycle (the next peak value). In free decaying vibration, the logarithm of the amplitude ratio reflects the magnitude of damping. Cracks increase the damping ratio by more than 100%.

[0171] Damping criterion function Determine if the damping ratio of the target washer is abnormal: ;in, To design a standard damping ratio (theoretical damping without defects) to serve as a benchmark for comparing measured damping ratios. Damping ratio compared to design standard The percentage of cases exceeding 100% indicates the presence of internal cracks (friction energy loss at the crack interface).

[0172] like If so, the damping ratio of the target washer is determined to be abnormal;

[0173] like If the damping ratio of the target washer is normal, then it is determined that the damping ratio is normal.

[0174] This embodiment also provides a method for obtaining sampling data, specifically:

[0175] Set the sampling size:

[0176] According to GB / T 2828.1-2012 standard, select an appropriate inspection level and AQL value. For example, for regular batches (N=1000-5000), use general inspection level II with a sampling quantity n of 80-200 pieces. ;in, To minimize the function and ensure that the sample size does not exceed the upper limit of 200 items, To maximize the function, ensure that the sample size is not less than the lower limit of 80 items. This indicates rounding down, discarding the decimal part of the calculation result. The total number of pieces in the batch (1000-5000 pieces) is used to calculate the sampling base.

[0177] Perform stratified sampling:

[0178] Stratified sampling is performed based on production time period, equipment channel, and material batch. For example, products produced every 2 hours constitute one sub-batch, with 20 pieces sampled from each channel, and 30 pieces sampled from each different raw material batch. ; ; ;in, This refers to the stratified sampling size based on time, such as drawing 10 items every 2 hours. This refers to the sampling number stratified by equipment channel, such as 20 items per channel. This refers to the stratified sampling number based on material batches, such as 30 pieces per batch. This indicates rounding down, discarding the decimal part of the calculation result. This represents the total number of pieces in the same batch, used as the base for calculating time stratification.

[0179] Surface defect verification was performed on the sampled metal gaskets:

[0180] The sampled parts underwent full-surface manual re-inspection. A 10× magnifying glass was used to examine for minute scratches and pits, while a stereomicroscope (20-50×) was used to observe the microstructure. Standard defect samples were compared to identify defects such as cracks and indentations. ;

[0181] Using higher-precision testing equipment, such as a 12-megapixel camera or 3D line laser scanning, for instrument re-inspection. If the high-precision equipment discovers defects that the original equipment missed, it is determined that the original equipment missed the detection. ;

[0182] Calculate the false negative rate: ;

[0183] Determining the false negative rate function To determine whether the vision system is reliable:

[0184] ;

[0185] like If so, the visual system is deemed unreliable;

[0186] like If so, the visual system is deemed reliable;

[0187] Internal defect inspection was performed on the sampled metal gaskets:

[0188] The testing method is selected based on the gasket type and risk level. For example, ordinary sealing gaskets are tested using ultrasonic testing (UT), automotive engine gaskets are tested using ultrasonic testing plus industrial CT testing, and aerospace gaskets are tested using industrial CT testing plus X-ray testing.

[0189] ;

[0190] Ultrasonic testing (UT) specifically involves using a portable ultrasonic flaw detector (frequency 5-10MHz) to perform C-scan and A-scan on the sampled metal gasket. The ultrasonic testing judgment function is then used to determine whether the ultrasonic testing data is normal, thereby identifying any internal defects.

[0191] ;

[0192] like If the ultrasonic test data is abnormal, it means that there is an internal defect.

[0193] like If the ultrasonic test data is normal, it means that no internal reflected wave was detected by ultrasonic waves. The sample is likely to have only surface defects and no internal abnormalities.

[0194] Specifically, industrial CT inspection involves using a microfocus X-ray CT system (resolution ≤ 5μm) to perform a 360° rotational scan on the sampled metal gasket. The CT inspection judgment function is then used to determine whether the industrial CT inspection data is normal, thereby identifying any internal defects. ;

[0195] like If so, the industrial CT inspection data is determined to be abnormal, meaning there is an internal defect;

[0196] like If the industrial CT test data is normal, it means that the sample has no internal holes or cracks and its internal quality is qualified.

[0197] Specifically, X-ray inspection involves using a 160kV microfocus X-ray machine to vertically irradiate the sampled metal gasket. The X-ray inspection data is then analyzed using a radiographic inspection judgment function to determine if the data is normal, thereby identifying any internal defects.

[0198] ;

[0199] like If so, the X-ray detection data is determined to be abnormal, meaning there is an internal defect;

[0200] like If the X-ray detection data is normal, it means that the sample has no internal density abnormalities and no internal defects.

[0201] The number of internally defective items is counted and recorded as follows: : ;in, This refers to the total number of items sampled, such as 100 items. This serves as a sampling index, used to iterate through all sampled samples and count the total number of defects. ;

[0202] Calculate the defect rate: ;

[0203] Data analysis and batch quality assessment were performed on the sampled metal gaskets.

[0204] Batch defect rate inference using hypergeometric distribution or binomial distribution: ;in, This is the most conservative estimate (i.e., the upper limit of the worst-case scenario) of the true defect rate of a batch at a given confidence level (e.g., 95%), used to adopt a conservative strategy in quality judgment, ensuring that the consumer risk (acceptance of non-conforming batches) is below 5%, thus protecting the interests of downstream customers. This indicates that, based on the sample defect rate obtained through sampling, calculated using a binomial distribution or hypergeometric distribution, the probability that the actual defect rate does not exceed this upper limit is 95%. If the defect rate of the entire batch is below the acceptable standard, the batch is deemed acceptable; otherwise, it is rejected to avoid quality risks.

[0205] Batch quality judgment function To determine whether to accept this batch of washers: ;in, The acceptable quality limit, such as 0.4%, is the maximum acceptable defect rate.

[0206] like If the batch meets the quality requirements, it can be accepted and proceed to the next process or delivered to the customer.

[0207] like If the batch fails to meet the quality standards, it will be rejected and must undergo 100% inspection, rework, or be scrapped and isolated.

[0208] Based on the sampling comprehensive judgment function Determine whether the sampled gaskets meet the normal requirements: ;in, ;

[0209] like If so, the sampled gaskets are deemed to be in normal condition;

[0210] like If so, it is determined that the sampled gaskets do not meet the normal requirements.

[0211] Example 3 is an improvement on Example 2. In this example, a preliminary misjudgment analysis is performed on the preliminary detection results, specifically as follows:

[0212] Washers that are deemed abnormal are classified as abnormal washers.

[0213] Place the faulty gasket back into the same testing equipment and repeat the test, such as 3-5 times consecutively. ;in, For the result of duplicate detection, This indicates that if more than 80% of the results are abnormal in 3-5 consecutive tests, it is determined not to be a false positive (the probability of a true defect is >90%). If the anomaly rate is less than 80% (fluctuations or all normal), it is determined to be a misjudgment (accidental interference or equipment problem). The threshold for judgment (80%) is set when the percentage of abnormal occurrences in repeated detections is ≥80%. In this case, the results are considered consistent, random errors are excluded, and the defect is determined to be a genuine defect. The total number of tests. Index for the number of detections;

[0214] The defective gasket was re-inspected using a separate testing system of a different brand and operating principle.

[0215] ;in, This indicates that the original equipment was determined to be abnormal. This indicates that the second device has been identified as abnormal. This indicates that the second device is considered normal. This indicates that the original equipment was determined to be normal. For cross-validation results from multiple devices, This indicates that both devices are malfunctioning, meaning the gasket has a genuine defect. This indicates that only the original equipment is malfunctioning, meaning the initial assessment of the gasket as malfunctioning was a misjudgment. The results are inconsistent and require further analysis (manual review or parameter verification).

[0216] The threshold checking function checks whether the thresholds are set abnormally, specifically whether the grayscale threshold, area threshold, and contrast threshold are too low, leading to an increase in false positives.

[0217] Adjust the threshold by 10%, such as increasing the grayscale threshold by 10-20, the area threshold by 0.1 mm², or the contrast threshold by 5-10%. ;in, The original threshold, The threshold value is adjusted. After adjusting the threshold, the washer is retested. If it returns to normal, it indicates that the original threshold was too strict, leading to a false positive. ;in, For the threshold check results, This indicates that the detection results returned to normal after the threshold was lowered by 10%, meaning that the abnormal judgment of the washer was a misjudgment (the original threshold was too strict). This indicates that the error persists even after the threshold callback, meaning that the abnormal gasket determination is not a misjudgment (the actual defect or threshold is not the main cause).

[0218] The feature extraction algorithm and illumination compensation algorithm are checked for anomalies using algorithm and model validation functions.

[0219] Slightly rotate the washer angle and re-detect. If the anomaly disappears, it indicates a misjudgment of lighting. Collect cases where normal textures were misjudged as defects into a dataset. Add the misjudged samples to the training set, retrain the model, and use the new model to detect historical misjudged samples. If the results improve, it indicates a texture misjudgment. Use techniques such as Grad-CAM to observe whether the model focuses on texture regions.

[0220] ;in, The results are the detection results after rotation. This indicates that the abnormality disappeared after rotating the washer, meaning the washer detection error was a misjudgment caused by lighting / angle. This indicates that the problem persists even after rotation, meaning the washer malfunction is not due to a misjudgment of the lighting. The results of texture misjudgment check. This indicates that the model focuses on non-defect texture areas, meaning that the abnormal gasket judgment is a misjudgment caused by texture recognition (normal processing texture is incorrectly identified as a defect). This indicates that the model focuses on the actual defect area, meaning that the abnormal gasket detection is not a texture misjudgment (the model correctly identifies the actual defect). The results validate the algorithm and model. This indicates that the gasket detection error was a misjudgment. This indicates that the abnormal gasket detection is not a misjudgment by the algorithm;

[0221] The vibration disturbance detection function is used to check for any abnormal vibrations in the workshop.

[0222] Re-inspect the abnormal gaskets while the machine is stopped or under vibration reduction. If they are normal, it indicates that the vibration caused the misjudgment. ;in, This is the vibration acceleration amplitude, 0.5 times the gravitational acceleration. When the workshop vibration exceeds this value, it may introduce detection noise, leading to misjudgment. For vibration interference detection results, This indicates that the vibration was greater than 0.5g and the machine returned to normal after shutdown, meaning the gasket malfunction was a misjudgment (caused by vibration). This indicates that the vibration is within the acceptable range or that the abnormality persists after the machine is stopped, meaning that the abnormality of the gasket is not a misjudgment of the vibration.

[0223] The light source stability check function is used to check for any abnormal fluctuations in illuminance.

[0224] If the standard grayscale value is less than 2 after 100 consecutive tests using a standard grayscale card, it indicates that the light source is stable; otherwise, it indicates that the light source is unstable. ;in, The standard deviation of the grayscale value. This indicates two gray levels. When the standard deviation of the gray value exceeds two gray levels, it indicates unstable lighting (flickering or fluctuations), resulting in poor image quality and potentially being misjudged as scratches or other defects. The results of the light source stability check are as follows. This indicates that the light source is unstable, meaning the abnormal gasket detection is a misjudgment (caused by the light source). This indicates that the light source is stable, meaning that the abnormal gasket detection is not a misjudgment.

[0225] The electromagnetic interference verification function is used to check for the presence of electromagnetic interference.

[0226] After isolating the testing area with a shield, the gasket is re-inspected. If the test is normal after shielding, it indicates the presence of electromagnetic interference. ;in, The results are after shielding. This indicates that the anomaly disappeared after being blocked. This indicates that the problem persists even after blocking. For electromagnetic interference verification results, This indicates the presence of electromagnetic interference, meaning the gasket malfunction was a misjudgment (caused by electromagnetic interference). This indicates that the abnormal gasket condition was not caused by electromagnetic interference.

[0227] Analyze parameters for incorrect judgments.

[0228] Specifically, the analysis of parameters for misjudgment during execution is as follows:

[0229] Plot the time / location distribution of defects within the batch;

[0230] Analysis of the defect distribution of abnormal washers:

[0231] If the defects are randomly distributed in the distribution map, it indicates that they are real defects; if the defects are concentrated in the distribution map, it indicates that it is a systematic misjudgment. ;

[0232] Verify the function using OC curves. To determine if the detection system has any abnormalities:

[0233] Comparing the actual detection rate with the theoretical value of the OC curve, if the actual false detection rate exceeds 2% and AQL = 0.65, it indicates that the detection system is inaccurate. The OC curve is a theoretical tool describing the theoretical acceptance probability of batches at different quality levels under a specific sampling plan.

[0234] ;in, The false detection rate threshold is defined as follows: when the actual false detection rate exceeds 2%, it indicates that the actual performance is worse than the theoretical value of the OC curve, and the detection system is considered to be out of acceptable range, requiring calibration or repair. When the actual false detection rate is below 2%, it indicates that the actual performance meets the theoretical expectations. The actual false positive rate is the proportion of qualified products that the detection system incorrectly identifies as unqualified. This represents an acceptable quality level and serves as the benchmark for batch quality judgment; batches below the AQL are accepted, while batches above the AQL are rejected.

[0235] like If the detection system fails to detect the error, it is determined that the detection system is malfunctioning.

[0236] like If so, the detection system is considered to be functioning normally;

[0237] Perform production record traceability to determine the completeness of production records and the stability of the production process:

[0238] Check whether the raw material batch has undergone UT / RT testing, whether the stamping die is worn, and whether the heat treatment process parameters deviate. If the production records are complete and the process is stable, it indicates that the probability of the anomaly being a real defect is low. ;in, For production record traceability results, This indicates incomplete records or unstable processes, meaning the probability of the gasket abnormality being a genuine defect is low. This indicates that the records are complete and the process is stable, meaning that the gasket abnormality is more likely to be a real defect.

[0239] Failure Mode Analysis Function To determine if there is a misjudgment of production actions:

[0240] If the anomaly type is indentation / white spot, trace back to the monitoring video of the previous process to confirm whether there was an actual pressing action. If there was no pressing process in the previous process, the possibility of misjudgment is high.

[0241] ;in, This indicates that there was no pressing action in the previous process, but the abnormality is an indentation, meaning the gasket abnormality was a misjudgment. This indicates that there was a pressing action in the previous process, meaning that the abnormal gasket may be a real defect.

[0242] Calculate the preliminary misjudgment comprehensive judgment result: ;

[0243] Based on the preliminary comprehensive judgment function To determine whether there was a misjudgment in the initial assessment:

[0244] ;

[0245] like If so, it is preliminarily determined that there is a misjudgment;

[0246] like Therefore, it can be preliminarily determined that there is no misjudgment.

[0247] This embodiment also provides defect analysis, specifically:

[0248] Based on the visual inspection results, analyze the surface defect morphology:

[0249] ;in, Based on morphological analysis, if the gasket exhibits cracks, dents, or deformation accompanied by cracking, it indicates potential internal defects requiring further analysis and testing. For example, if the crack edge is V-shaped or wedge-shaped, with a large opening and rough edges, it may extend inwards. If the edge of the dented area shows material tearing or fracture characteristics rather than simple indentation, it may suggest internal defects. Overall deformation of the gasket with surface cracks may indicate a connected defect caused by internal stress. The presence of deep-penetrating cracks or irregular pits may indicate internal defects. This indicates that the surface defect morphology is normal and it is most likely just a surface defect.

[0250] Three-dimensional depth information analysis of surface defects:

[0251] Depth information is obtained using structured light or 3D vision technology. Striped light is projected onto the gasket surface, and the degree of stripe distortion is analyzed to estimate the defect depth. If the defect depth exceeds 30%-50% of the gasket thickness, it is highly likely to extend into the interior, meaning there may be a defect inside the gasket, requiring further analysis and inspection. ;

[0252] in, For in-depth analysis results, This indicates that surface defects in the gasket may extend into the interior, meaning there may be defects inside the gasket. This indicates that the gasket likely only has surface defects.

[0253] The abnormal gasket is inspected using ultrasound to determine whether the ultrasonic data of the abnormal gasket is abnormal:

[0254] High-frequency ultrasound propagates inside the metal. If it encounters a defect, it generates a reflected wave. If a defect is visually detected but the ultrasound does not detect any abnormality, it indicates that the defect is only on the surface. If a defect is visually detected and the ultrasound detects a reflected wave at the corresponding location, it indicates that defects exist both internally and on the surface. ;in, The results are from the ultrasonic testing. This indicates that the ultrasonic waves detected internal reflected waves, meaning that there are defects both inside and on the surface of the gasket. This indicates that no abnormality was detected by ultrasonic testing, meaning the gasket likely only has surface defects.

[0255] The abnormal gasket is inspected using X-rays to determine whether the X-ray data of the abnormal gasket is abnormal:

[0256] X-rays penetrate metal, forming a shadow image on film or a detector. Areas of density variation within this image represent defects. If the X-ray image shows a defect only near the surface, it indicates a surface defect. If the X-ray image shows a defect extending from the surface into the interior, it indicates a connected defect. ;in, The results are from X-ray examination. This indicates that the X-ray shows a connection defect between the inside and outside of the gasket, meaning the gasket has both internal and surface defects. This indicates that the X-ray shows only surface defects, meaning the gasket has only surface defects.

[0257] By performing magnetic particle testing on the defective gaskets, it can be determined whether the magnetic particle data of the defective gaskets is abnormal.

[0258] After magnetizing the washer, surface and near-surface defects can cause magnetic field leakage, attracting magnetic powder and forming visible traces. If the defect is only visually detected but no abnormality is found in the magnetic powder test, it indicates that it is only a surface defect (depth < 2mm). If the defect is detected by both visual inspection and magnetic powder testing, it indicates a near-surface or internal / external connected defect. ;in, These are the results of magnetic particle testing. This indicates that a defect was detected by magnetic particle testing, namely a defect near the surface of the washer or a defect that connects the inside and outside. This indicates that the magnetic particle testing showed no abnormalities, meaning the gasket only has surface defects.

[0259] Comprehensive assessment of execution deficiencies.

[0260] The comprehensive assessment of execution deficiencies includes:

[0261] Assess the impact of defects on service life:

[0262] The impact on service life is assessed based on the type and size of the defect. For cracks, if the length exceeds 10% of the washer diameter or the depth exceeds 20% of the thickness, the service life is severely affected. For holes, if the diameter exceeds 30% of the washer thickness, it may lead to premature failure.

[0263] ;

[0264] in, For the service life impact assessment results, This indicates that the crack or hole size exceeds the standard, meaning that the defect in the gasket seriously affects its service life. This indicates that the dimensions are within specifications, meaning the defect in the washer will not affect its service life. This refers to the washer diameter, used to assess whether the crack length exceeds the limit. The gasket thickness is used to assess whether the hole diameter or crack depth exceeds the standard.

[0265] Assess the impact of the defect on normal use:

[0266] Assess the impact on normal use based on the location and type of the defect. For surface cracks, if located on the sealing surface, they may lead to leakage and affect normal use. For internal holes, if near stress concentration areas, they may lead to premature fracture.

[0267] ;

[0268] in, The impact assessment results are for normal use. This indicates that the crack is located on the sealing surface or the hole is near the stress concentration area, meaning that the defect in the gasket affects normal use. This indicates that the defect location is safe, meaning the defect in the gasket does not affect normal use;

[0269] The comprehensive conclusion of the calculation defects is denoted as: : ;

[0270] Defect comprehensive judgment function To determine whether the defective gasket is ultimately qualified: ;

[0271] like If so, the abnormal gasket is deemed to be ultimately qualified;

[0272] like If the gasket is found to be defective, it will be deemed unqualified.

[0273] This embodiment achieves closed-loop management from surface inspection to internal defect verification, and from single-piece analysis to batch statistics, ensuring "zero omissions" of defects. It replaces the blindness of traditional manual sampling inspection, transforming quality risk from "post-event discovery" to "pre-event prevention." Through a five-layer misjudgment analysis mechanism (re-inspection, equipment comparison, parameter verification, environmental investigation, and batch statistics), it reduces the misjudgment rate, minimizes waste of qualified products, lowers enterprise quality costs, and enhances the reliability of the inspection system. By analyzing internal defects through macroscopic parameters (weight, center of gravity, stiffness), it improves the detectability of internal defects, quickly screening out internal holes, cracks, and other problems that are missed by traditional visual inspection, avoiding early failure of metal gaskets after installation. It automatically matches the inspection intensity (sampling / full inspection) according to the gasket application scenario (general / automotive / aerospace), reducing inspection costs and avoiding "one-size-fits-all" over-inspection, achieving optimal allocation of inspection resources. It establishes a standard defect library, OC curve monitoring, and TPR / FPR capability index tracking, forming a quality data closed loop, providing data support for process optimization, and driving continuous improvement in the accuracy of the inspection system.

Claims

1. A fully automated inspection method for metal washers based on industrial vision, characterized in that: include: Obtain the target washer; By inspecting the target gasket using industrial vision, a preliminary judgment is made as to whether the target gasket is qualified, and a preliminary inspection conclusion is formed. If the target gasket is initially determined to be qualified, a second analysis will be conducted. If the target gasket is initially determined to be unqualified, a preliminary misjudgment analysis should be conducted on the preliminary test results. If the initial assessment indicates a misjudgment, a second analysis will be conducted. If the initial assessment determines that there is no misjudgment, then a defect analysis will be conducted.

2. The fully automated inspection method for metal washers based on industrial vision according to claim 1, characterized in that: Industrial vision is used to inspect the target gasket and make a preliminary judgment on whether the target gasket is qualified. Specifically: Acquire an image of the target washer and designate it as the target image; Preprocess the target image; Geometrically locate the target image; Extract defect features; Calculate the overall defect score for each candidate region; By region determination function For each candidate region, determine whether it has a defect; like If so, the candidate region is determined to be defective; like If so, the candidate region is determined to be free of defects; Preliminary defect judgment function A preliminary assessment is made to determine whether the target gasket is qualified. like If so, the target washer is preliminarily determined to be qualified; like If so, it is preliminarily determined that the target washer is unqualified.

3. The fully automated metal gasket inspection method based on industrial vision according to claim 1, characterized in that: A second analysis is conducted, specifically as follows: Perform anomaly analysis on the target gasket and generate anomaly judgment data; Obtain sampling data; Calculate the multi-parameter comprehensive score of the target washer; Through the exception detection function Determine if there are any abnormalities inside the target gasket; like If so, it is determined that there is an abnormality inside the target washer; like If so, it is determined that there is no abnormality inside the target washer; By comprehensive judgment function A comprehensive judgment is made as to whether the target gasket is qualified; like If so, the target washer is ultimately deemed qualified; like If so, then defect analysis will be performed.

4. The fully automated metal gasket inspection method based on industrial vision according to claim 3, characterized in that: Anomaly analysis is performed on the target gasket to generate anomaly judgment data, specifically: Collect the weight of the target washer; Calculate the relative deviation value; Weight determination function To determine if the weight of the target washer is abnormal; like If so, the weight of the target washer is determined to be abnormal; like If the weight of the target washer is normal, then it is determined that the weight of the target washer is normal. Collect the center of gravity position of the target washer; Calculate the center of gravity offset; Using the centroid determination function To determine whether the center of gravity of the target washer is abnormal; like If so, the center of gravity of the target washer is determined to be abnormal; like If the target washer's center of gravity is normal, then it is determined that the target washer's center of gravity is normal. Collect the compressive stiffness of the target washer; Stiffness determination function To determine whether the stiffness of the target washer is abnormal; like If so, the stiffness of the target washer is determined to be abnormal; like If the stiffness of the target washer is normal, then it is determined that the stiffness of the target washer is normal. Collect the natural frequency of the target washer; Frequency determination function To determine whether the natural frequency of the target washer is abnormal; like If so, the natural frequency of the target washer is determined to be abnormal; like If the natural frequency of the target washer is normal, then it is determined that the natural frequency of the target washer is normal. Collect the damping ratio of the target washer; Damping criterion function To determine whether the damping ratio of the target washer is abnormal; like If so, the damping ratio of the target washer is determined to be abnormal; like If the damping ratio of the target washer is normal, then it is determined that the damping ratio is normal.

5. The fully automated inspection method for metal washers based on industrial vision according to claim 3, characterized in that: The sampling data is obtained as follows: Set the sampling size; Perform stratified sampling; Surface defects were verified on the sampled metal gaskets; Internal defect detection was performed on the sampled metal gaskets; Data analysis and batch quality assessment were performed on the sampled metal gaskets. Based on the sampling comprehensive judgment function To determine whether the sampled gaskets meet the normal requirements; like If so, the sampled gaskets are deemed to be in normal condition; like If so, it is determined that the sampled gaskets do not meet the normal requirements.

6. The fully automated inspection method for metal washers based on industrial vision according to claim 1, characterized in that: A preliminary misjudgment analysis was conducted on the preliminary test results, specifically as follows: Washers that are deemed abnormal are classified as abnormal washers. The faulty gasket was placed back into the same testing equipment for repeated testing; The defective gaskets were re-inspected using a separate testing system. The threshold check function is used to check whether the threshold setting is abnormal. The algorithm and model validation function is used to check for anomalies in the feature extraction algorithm and the illumination compensation algorithm. The vibration interference detection function is used to check whether there are any abnormalities in the workshop vibration. The light source stability check function is used to check for any abnormalities in illuminance fluctuations. The electromagnetic interference verification function is used to check for the presence of electromagnetic interference. Analyze parameters for incorrect judgments.

7. The fully automated metal gasket inspection method based on industrial vision according to claim 6, characterized in that: Perform a misjudgment parameter analysis, specifically as follows: Plot the time / location distribution of defects within the batch; Analyze the defect distribution of abnormal washers; Verify the function using OC curves. To determine if there are any abnormalities in the detection system; like If so, it is determined that the detection system is abnormal; like If so, the detection system is considered to be functioning normally; To trace production records and determine whether the records are complete and whether the production process is stable; Failure Mode Analysis Function To determine whether there is a misjudgment of production actions; Calculate the preliminary misjudgment comprehensive judgment result; Based on the preliminary comprehensive judgment function To determine whether there was a misjudgment in the initial assessment; like If so, it is preliminarily determined that there is a misjudgment; like Therefore, it can be preliminarily determined that there is no misjudgment.

8. The fully automated inspection method for metal washers based on industrial vision according to claim 3, characterized in that: Defect analysis is performed, specifically as follows: Analyze the morphology of surface defects based on the visual inspection results; Perform three-dimensional depth information analysis on surface defects; Abnormal gaskets are detected by ultrasonic waves to determine whether the ultrasonic data of the abnormal gaskets is abnormal. The abnormal gasket is inspected using X-rays to determine whether the X-ray data of the abnormal gasket is abnormal. By performing magnetic particle testing on abnormal washers, it can be determined whether the magnetic particle data of the abnormal washers is abnormal. Comprehensive assessment of execution deficiencies.

9. The fully automated inspection method for metal washers based on industrial vision according to claim 8, characterized in that: A comprehensive assessment of the implementation deficiencies is conducted, specifically as follows: Assess the impact of defects on service life; Assess the impact of the defect on normal use; Comprehensive conclusions regarding computational defects; Defect comprehensive judgment function To determine whether the defective gasket is ultimately qualified; like If so, the abnormal gasket is deemed to be ultimately qualified; like If the gasket is found to be defective, it will be deemed unqualified.

Images