Nuclear power sealing gasket production quality detection method based on machine vision

Abstract

The invention discloses a nuclear power sealing gasket production quality detection method based on machine vision, and relates to the technical field of image data processing.The method comprises the steps that firstly, a high-precision morphology mapping matrix is established through multispectral imaging and structured light scanning, and digital characterization of surface flaws is achieved; a texture distortion rate and edge gradient consistency parameter is extracted through a visual analysis algorithm, and a standardized surface quality evaluation index is generated; a material physical mechanism model is introduced to associate with hot working historical data, the internal structure uniformity probability is calculated through inversion, and a defect prediction index is generated; and finally, constructing a multi-dimensional evaluation matrix coupling dominant risk and implicit failure components, and calculating a comprehensive reliability coefficient to judge a grade. The method has the advantages that manual subjective experience interference is eliminated, linkage evaluation of surface process defects and internal microscopic internal injuries is achieved, performance degradation of the gasket in nuclear power high-temperature and high-pressure and intense radiation environments is predicted, long-acting reliability of a sealing element is guaranteed, and the safety target of zero leakage is guaranteed.

Description

Technical Field

[0001] This invention relates to the field of image data processing technology, specifically to a machine vision-based method for quality inspection of nuclear power plant sealing gaskets. Background Technology

[0002] Nuclear power plant gaskets are key components ensuring the safe operation of nuclear power plants. Their development originated from the nuclear industry's stringent requirements for fluid sealing under extreme operating conditions. As global nuclear power technology has evolved from early reactor types to third and fourth generation, gasket materials have progressed from single metals to nuclear-grade high-purity graphite and complex composite structures. They are widely used in nuclear power plant reactor pressure vessels, steam generators, main pumps, and various nuclear-grade valve systems, serving as the core line of defense for maintaining system tightness and preventing the leakage of radioactive materials.

[0003] Due to their long-term exposure to extreme environments of high temperature, high pressure, and strong radiation, any quality defect can pose a serious safety risk. Therefore, quality inspection during the production process is crucial. It has evolved from traditional manual inspection to an intelligent inspection system integrating high-precision dimensional measurement, non-destructive testing, and material performance analysis to ensure the "zero leakage" safety goal. However, a common drawback in the industry is the excessive reliance on human experience in the inspection process, leading to insufficient standardization. Traditional quality inspection often employs visual inspection or semi-automated physical measurement. Because nuclear-grade gaskets have extremely high requirements for surface flatness and minute defects, differences in subjective judgment standards among different inspectors, as well as visual fatigue from prolonged work, can easily lead to the omission of minute cracks or surface deformations, making it difficult to guarantee absolute consistency of inspection results under mass production.

[0004] In addition, a closely related technical drawback is that macroscopic surface inspection cannot characterize the uniformity of the material's internal microstructure and its response to extreme operating conditions in real time. Because the aforementioned surface inspection methods only address "visible defects," existing inspection technologies often overlook internal lattice defects or microscopic compositional segregation caused by thermal stress during manufacturing. These "internal injuries" may appear normal at room temperature, but when combined with surface process defects, they can lead to unexpected performance degradation of the gasket under the high temperature, high pressure, and strong radiation environment of nuclear power plants. This connection between the limitations of surface inspection and the lack of assessment of internal microscopic integrity is a core pain point in ensuring the long-term reliability of sealing components. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a machine vision-based method for quality inspection of nuclear power plant sealing gaskets, thereby overcoming the aforementioned technical deficiencies.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a machine vision-based method for quality inspection of nuclear power plant sealing gaskets, comprising the following steps:

[0007] S1. Use a multispectral imaging device to acquire macroscopic surface image data of the nuclear power plant sealing gasket to be tested, and simultaneously establish a high-precision surface morphology mapping matrix to realize the digital characterization of the gasket surface geometry and minor defects.

[0008] S2. Using a surface stress visual analysis algorithm, extract the texture distortion rate and edge gradient consistency parameters from the macroscopic surface image data to quantitatively characterize the processing consistency of the gasket surface and generate surface quality evaluation indicators.

[0009] S3. Using a preset material physics mechanism model, the surface morphology mapping matrix is ​​correlated with the historical data of heat treatment in the production process to inversely calculate the probability of uniformity of microstructure inside the material and generate internal defect prediction index.

[0010] S4. The surface quality evaluation index and the internal defect prediction index are weighted and fused to construct a multi-dimensional quality assessment matrix, and the comprehensive reliability coefficient is calculated.

[0011] S5. Based on the comparison result between the comprehensive reliability coefficient and the preset safety threshold, the gasket quality level is automatically output and fed back to the production execution system for classification and interception or process parameter compensation.

[0012] Preferably, in step S1, a multi-axis industrial camera deployed above the production line collects reflected light intensity data from multiple angle light sources, and structured light scanning technology is used to obtain spatial depth information of each coordinate point on the gasket surface, thereby generating a high-precision surface topography mapping matrix.

[0013] Preferably, in step S2, the regularity of surface texture is analyzed by calculating the gray-level co-occurrence matrix between adjacent pixels in the image, and micron-level cracks and indentations are identified using an edge detection operator. The identified defect area, depth, and distribution density are used as variables, and after dimensionless processing, surface quality evaluation indicators are obtained.

[0014] Preferably, in step S3, the material physical mechanism model is based on the thermoelastic mechanics equation. It extracts the micro-undulation features in the surface morphology as the cause of thermal stress concentration, and combines the historical data of compositional segregation of the batch of metal materials to assess the risk of internal lattice defects caused by thermal stress.

[0015] Preferably, in step S4, the calculation logic of the comprehensive reliability coefficient is as follows: the surface quality evaluation index is defined as the explicit risk component, the internal defect prediction index is defined as the implicit failure component, the coupling correlation value between the two is calculated through a nonlinear regression function, and a performance degradation correction factor under extreme working conditions is introduced to finally obtain the comprehensive reliability coefficient reflecting the long-term sealing capability.

[0016] Preferably, in step S5, the criteria for automatically outputting the quality grade of the gasket are as follows: when the comprehensive reliability coefficient is in the first range, it is marked as a nuclear-grade qualified product; when only the surface indicators are qualified but the internal predicted indicators are abnormal, it is marked as a downgraded product or a product subject to mandatory re-inspection; when there are obvious defects on the surface and the internal risk is high, it is marked as a scrapped product.

[0017] Preferably, the process of generating the surface quality evaluation index further includes: performing a Fourier transform on the collected light intensity distribution, extracting high-frequency components in the frequency domain to identify surface micro-irregularities, and comparing them with the spectral characteristics of a standard sample to eliminate interference from human experience under different testing environments.

[0018] Preferably, the calculation logic of the internal defect prediction index is as follows: establish the mapping relationship between surface geometrical abrupt changes and internal stress field distribution, calculate the internal stress concentration coefficient caused by surface micro-deformation, and multiply the coefficient by the radiation embrittlement index of the material to obtain a prediction index characterizing the failure risk of the gasket in a strong radiation environment.

[0019] Preferably, S5 further includes establishing a closed-loop quality traceability logic, which associates and stores the comprehensive reliability coefficient of each gasket with its unique production serial number, raw material furnace number and visual inspection image sequence to ensure the traceability of the data source.

[0020] Preferably, the calculation of the coupling correlation value is achieved by analyzing the superposition effect of surface process defects and internal component segregation under high temperature and high pressure environment. If the internal structure uniformity corresponding to the area where the surface microcrack is located is lower than the preset uniformity threshold, the deduction weight of the comprehensive reliability coefficient is increased exponentially.

[0021] This invention provides a machine vision-based method for quality inspection of nuclear power plant sealing gaskets during production. It offers the following advantages:

[0022] (1) This method for quality inspection of nuclear power sealing gaskets based on machine vision acquires reflected light intensity data and spatial depth information from multiple angle light sources by using a multi-axis industrial camera and a multi-spectral imaging device, and simultaneously establishes a high-precision surface morphology mapping matrix, thereby achieving high-precision digital characterization of the surface geometry and minor defects of nuclear power sealing gaskets; combined with texture distortion rate analysis based on gray-level co-occurrence matrix, and using edge detection operators to accurately identify micron-level cracks and indentations, the identified defect area, depth and distribution density are used as variables to generate standardized surface quality evaluation indicators; combined with the high-frequency components extracted by Fourier transform to identify surface micro-unevenness, and comparing it with the spectral characteristics of standard samples, this technology effectively eliminates the risk of missed detection caused by subjective judgment differences and visual fatigue in traditional manual visual inspection, and fundamentally solves the industry pain points of insufficient standardization of nuclear-grade gasket inspection and difficulty in guaranteeing absolute consistency of inspection results in large-scale production.

[0023] (2) This machine vision-based method for quality inspection of nuclear power sealing gaskets further introduces a material physics mechanism model based on thermoelastic mechanics equations, deeply correlates the surface morphology mapping matrix with historical data of thermal processing and segregation data of metal material composition during the production process, inversely calculates the probability of internal structure uniformity and generates internal defect prediction index; by calculating the internal stress concentration coefficient caused by surface micro deformation and introducing radiation embrittlement index and performance decay correction factor, real-time characterization and extreme working condition response prediction of lattice defects and other hidden "internal damage" caused by thermal stress during processing are realized; by constructing a multi-dimensional quality assessment matrix containing explicit risk components and implicit failure components, and weighting and fusing surface process defects and internal structure uniformity with exponential weights, the comprehensive reliability coefficient obtained by this scheme can accurately assess the superposition effect of surface defects and internal micro damage, effectively solves the pain point of correlation between macroscopic detection limitations and lack of internal micro assessment, and ensures the long-term reliability of sealing components in the high temperature, high pressure and strong radiation environment of nuclear power. Attached Figure Description

[0024] Figure 1 This is a schematic diagram of the detection process in Example 1 of the machine vision-based nuclear power plant gasket production quality inspection method of the present invention;

[0025] Figure 2 This is a schematic diagram of the detection process in Example 2 of the machine vision-based nuclear power plant gasket production quality inspection method of the present invention. Detailed Implementation

[0026] 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.

[0027] Example 1:

[0028] Please see Figure 1 This invention provides a machine vision-based method for quality inspection of nuclear power plant sealing gaskets, comprising the following steps:

[0029] S1. Use a multispectral imaging device to acquire macroscopic surface image data of the nuclear power plant sealing gasket to be tested, and simultaneously establish a high-precision surface morphology mapping matrix to realize the digital characterization of the gasket surface geometry and minor defects.

[0030] S2. Using a surface stress visual analysis algorithm, extract the texture distortion rate and edge gradient consistency parameters from the macroscopic surface image data to quantitatively characterize the processing consistency of the gasket surface and generate surface quality evaluation indicators.

[0031] S3. Using a preset material physics mechanism model, the surface morphology mapping matrix is ​​correlated with the historical data of heat treatment in the production process to inversely calculate the probability of uniformity of microstructure inside the material and generate internal defect prediction index.

[0032] S4. The surface quality evaluation index and the internal defect prediction index are weighted and fused to construct a multi-dimensional quality assessment matrix, and the comprehensive reliability coefficient is calculated.

[0033] S5. Based on the comparison result between the comprehensive reliability coefficient and the preset safety threshold, the gasket quality level is automatically output and fed back to the production execution system for classification and interception or process parameter compensation.

[0034] In this embodiment, macroscopic surface image data is acquired by using a multispectral imaging device in S1 and a high-precision surface morphology mapping matrix is ​​established simultaneously, thereby realizing the digital characterization of the surface geometry and minor defects of the gasket, and providing a definite physical data basis for subsequent high-precision detection.

[0035] By executing the surface stress visual analysis algorithm of S2, the texture distortion rate and edge gradient consistency parameters are accurately extracted and surface quality evaluation indexes are generated, thereby realizing the quantitative evaluation of the consistency of processing technology and effectively avoiding the risk of missed detection and misjudgment in traditional manual inspection.

[0036] By introducing a material physics mechanism model through S3 and associating the morphology matrix with historical data of thermal processing, the probability of microstructure uniformity is calculated by inversion and an internal defect prediction index is generated. This completes the deep mapping from surface visual information to the health status of the material's internal structure, significantly enhancing the early warning capability for hidden damage under strong radiation conditions.

[0037] By using S4 to perform multi-dimensional weighted fusion of surface quality evaluation indicators and internal defect prediction indicators, a multi-dimensional quality assessment matrix is ​​constructed and a comprehensive reliability coefficient is calculated. This achieves a single-dimensional risk measurement of the overall service performance of the gasket, ensuring the systematicness and rigor of the evaluation conclusions.

[0038] By comparing the comprehensive reliability coefficient with the preset safety threshold in real time using S5, the gasket quality level is automatically output and fed back to the production execution system to execute classification interception or process parameter compensation instructions. This has built an intelligent closed-loop quality control system for the entire nuclear power sealing component production chain, which significantly improves the product warehousing qualification rate while ensuring the long-term service safety of nuclear-grade key components under extreme high temperature and high pressure conditions.

[0039] Example 2:

[0040] Please see Figure 2 In step S1, a multi-axis industrial camera deployed above the production line collects reflected light intensity data from multiple angle light sources, and structured light scanning technology is used to obtain spatial depth information of each coordinate point on the gasket surface, thereby generating a high-precision surface topography mapping matrix.

[0041] In step S2, the regularity of surface texture is analyzed by calculating the gray-level co-occurrence matrix between adjacent pixels in the image, and micron-level cracks and indentations are identified using an edge detection operator. The area, depth, and distribution density of the identified defects are used as variables, and after dimensionless processing, surface quality evaluation indicators are obtained.

[0042] In S3, the material physical mechanism model is based on the thermoelastic mechanics equation. It extracts the micro-undulation features in the surface morphology as the cause of thermal stress concentration, and combines the historical data of compositional segregation of this batch of metal materials to assess the risk of internal lattice defects caused by thermal stress.

[0043] In S4, the calculation logic of the comprehensive reliability coefficient is as follows: the surface quality evaluation index is defined as the explicit risk component, the internal defect prediction index is defined as the implicit failure component, the coupling correlation value between the two is calculated through a nonlinear regression function, and the performance degradation correction factor under extreme working conditions is introduced to finally obtain the comprehensive reliability coefficient reflecting the long-term sealing capability.

[0044] In S5, the criteria for automatically outputting the quality level of the gasket are as follows: when the comprehensive reliability coefficient is in the first range, it is marked as a nuclear-grade qualified product; when only the surface indicators are qualified but the internal predicted indicators are abnormal, it is marked as a downgraded product or a product subject to mandatory re-inspection; when there are obvious defects on the surface and the internal risk is high, it is marked as a scrapped product.

[0045] Furthermore, the integrated reliability coefficient is output through the integrated control module, and its value range is defined as a closed interval [0, 1]. When the integrated reliability coefficient approaches the maximum value of 1, it indicates that the physical state of the nuclear power plant sealing gasket under test is in the ideal benchmark model, that is, the texture distortion rate and edge gradient consistency of the gasket surface approach zero deviation, and the probability of internal microstructure uniformity is high, representing that the expected failure rate of the gasket under extreme operating conditions is at the theoretical minimum level; when the integrated reliability coefficient approaches the minimum value of 0, it indicates that there is a high coupling between explicit defects and latent damage in the gasket, indicating that the gasket has unstable characteristics before it is put into service. This quantitative trend transforms the complex physical quality evaluation into a single-dimensional risk measurement, realizing the transformation of the detection logic from simple visual feature extraction to multi-physics performance prediction.

[0046] Furthermore, the algorithm sets the texture distortion rate to be negatively correlated with the overall reliability coefficient to digitally characterize the impact of surface microchannel formation on the sealing pressure distribution; simultaneously, it sets the microstructure uniformity probability to be positively correlated, using a thermoelastic mechanical model to quantify the stability of the material's internal lattice structure. In addition, by introducing an extreme operating condition correction factor, a nonlinear regression algorithm is used to analyze the superimposed effect of surface process defects and internal component segregation under high temperature and pressure. This correction factor effectively simulates the amplification process of the stress intensity factor, ensuring accurate characterization of the gasket's performance degradation trend under strong radiation conditions within a normal temperature testing environment, thus resolving the technical logic defect of the disconnect between surface inspection and actual service performance evaluation.

[0047] To ensure the objectivity of the interval division, an objective threshold derivation methodology was adopted. The predicted leakage rate of the sealing surface was selected as a verifiable performance indicator, and a functional correlation model between the comprehensive reliability coefficient and the predicted leakage rate was established. By performing curvature-based second derivative analysis on this correlation function, key inflection points on the function curve were mathematically identified. The calculation process shows that significant abrupt changes in the curve curvature occur at comprehensive reliability coefficients of 0.90 and 0.70. These two inflection points define the physical boundaries of the material properties transitioning from a stable state to an accelerated degradation state. This threshold determination process eliminates human intervention and provides scientifically grounded classification logic support for production execution.

[0048] Based on the aforementioned mathematical identification inflection point, this embodiment divides the gasket quality into three application ranges with clear technical foundations. When the comprehensive reliability coefficient is in the range of [0.90, 1.00], it is defined as the nuclear-grade qualified range. Within this range, the predicted leakage rate remains below the preset safety benchmark, and the product is judged to be a nuclear-grade qualified product and automatically put into storage. When the comprehensive reliability coefficient is in the range of [0.70, 0.90), it is defined as the monitoring degradation range. It is determined that the gasket has a potential failure risk under extreme operating conditions, and degradation or mandatory physical re-inspection is performed. When the comprehensive reliability coefficient is in the range of [0.00, 0.70), it is defined as the high-risk failure range. The gasket is immediately intercepted through production execution, and while being marked for scrap, the process parameter compensation instruction is fed back to the front-end heat treatment unit, thereby realizing closed-loop real-time control of production quality.

[0049] The process of generating the surface quality evaluation index also includes: performing a Fourier transform on the collected light intensity distribution, extracting high-frequency components in the frequency domain to identify surface micro-irregularities, and comparing them with the spectral characteristics of standard samples to eliminate interference from human experience under different testing environments.

[0050] The calculation logic of the internal defect prediction index is as follows: establish the mapping relationship between surface geometrical abrupt changes and internal stress field distribution, calculate the internal stress concentration factor caused by surface micro-deformation, and multiply the factor by the radiation embrittlement index of the material to obtain the prediction index characterizing the failure risk of the gasket in a strong radiation environment.

[0051] The S5 also includes establishing a closed-loop quality traceability logic, which associates and stores the comprehensive reliability coefficient of each gasket with its unique production serial number, raw material furnace number and visual inspection image sequence to ensure the traceability of the data source.

[0052] The calculation of the coupling correlation value is achieved by analyzing the superposition effect of surface process defects and internal component segregation under high temperature and high pressure. If the internal structure uniformity corresponding to the area where the surface microcrack is located is lower than the preset uniformity threshold, the deduction weight of the comprehensive reliability coefficient is increased exponentially.

[0053] In this embodiment, multi-axis industrial cameras are deployed to collect reflected light intensity data from multiple angle light sources, and structured light scanning technology is combined to obtain spatial depth information. This ensures that the high-precision surface topography mapping matrix can completely reconstruct the spatial three-dimensional physical morphology of the gasket, providing a reference coordinate support for the subsequent quantitative characterization of defect depth. The surface texture regularity is analyzed using a gray-level co-occurrence matrix, and edge detection operators are combined to accurately identify the defect area, depth, and distribution density of micron-level cracks and indentations. At the same time, Fourier transform is performed on the light intensity distribution to extract high-frequency components in the frequency domain, thereby scientifically characterizing the surface micro-irregularity and eliminating the interference of human experience under different detection environments, thus improving the objectivity and robustness of the surface quality evaluation index.

[0054] Based on the thermoelastic mechanics equation, the micro-undulation features in the surface morphology are extracted as the cause of thermal stress concentration. This is coupled with the historical data of metal material composition segregation to establish the mapping relationship between surface geometrical abrupt changes and internal stress field distribution. By calculating the internal stress concentration coefficient and multiplying it with the radiation embrittlement index of the material, the internal defect prediction index characterizing the failure risk of the gasket under strong radiation environment is obtained, realizing the in-depth analysis and quantitative prediction of the hidden physical damage of the gasket.

[0055] The coupling correlation value between surface quality evaluation index and internal defect prediction index is calculated by nonlinear regression function, and a performance degradation correction factor under extreme working conditions is introduced to obtain the comprehensive reliability coefficient. In particular, an exponential deduction weight is applied to the region where the internal structure uniformity is lower than the preset uniformity threshold corresponding to the surface microcrack area. This accurately simulates the superposition effect of process defects and component segregation under high temperature and high pressure environment, and solves the technical bottleneck of the disconnect between surface information and actual service performance evaluation in conventional testing.

[0056] Finally, based on the comprehensive reliability coefficient, classification criteria for nuclear-grade qualification, downgraded use, mandatory re-inspection, and scrap interception were established. At the same time, a closed-loop quality traceability logic covering production serial number, raw material furnace number, and visual inspection image sequence was established to ensure that the quality data of each nuclear power sealing gasket can be traced back to the source of raw materials and production process. This not only improved the detection accuracy at the technical level through the fusion of multi-physical field features, but also built a comprehensive nuclear-grade product quality assurance system at the management level.

[0057] Furthermore, the control system generates a failure risk prediction index, hereinafter referred to as Ifail, through a computing unit. The value range of this index is limited to the closed interval [0, 1]. When the failure risk prediction index approaches the theoretical minimum value of 0, it indicates that the amplitude of the high-frequency component of the nuclear power plant sealing gasket under test is within a preset reference range, and the internal stress concentration coefficient approaches 1, representing that the gasket has structural integrity and long-term service potential under strong radiation environment. When the failure risk prediction index approaches the theoretical maximum value of 1, it indicates that the stress concentration effect caused by the surface geometric change and the radiation embrittlement characteristics of the material are nonlinearly superimposed, indicating that the gasket has a physical tendency to brittle fracture or through-leakage under high temperature and high pressure conditions. This index, by quantifying the coupling relationship between micromorphology and internal structure, realizes a technical leap from single visual defect identification to multi-physics field failure risk assessment.

[0058] In the specific parameter calculation logic, the image processing module performs a Fourier transform on the acquired light intensity distribution signal, extracts the amplitude of high-frequency components in the frequency domain to characterize the surface micro-irregularity, and sets it to be positively correlated with the failure risk prediction index, thereby quantifying the inducing effect of micro-undulations on local contact stress and eliminating the interference of human experience caused by fluctuations in the detection environment. Simultaneously, the mapping module establishes a topological mapping relationship between surface geometrical abrupt changes and the internal stress field, and multiplies the calculated internal stress concentration coefficient as a weighting factor with the material's radiation embrittlement index. In addition, the coupled calculation module analyzes the overlap between the surface microcrack region and the internal component segregation. If the internal microstructure homogeneity is found to be lower than a preset threshold, the deduction weight of the comprehensive reliability coefficient is increased exponentially, thereby sensitively characterizing the failure risk caused by the interface stress-irradiation synergistic effect in a strong radiation environment.

[0059] To ensure the scientific validity of the quality assessment criteria, this embodiment establishes the application range boundary of the failure risk prediction index through an objective threshold derivation method. The calculation unit establishes a mathematical mapping function between the failure risk prediction index and the critical crack propagation rate (da / dN), and uses first-order derivative analysis to identify the abrupt slope changes in the function curve, thus determining the physical inflection point where material properties transition from the steady-state to the unstable propagation stage. Based on simulation calculations and first-order derivative identification results, the system divides the application range of the failure risk prediction index into three discrete regions with clear physical meaning. The boundary of each region is determined by the abrupt change characteristics of the crack propagation rate, rather than being arbitrarily set.

[0060] Based on the identified physical inflection points, this embodiment executes the following classification and judgment logic: When the failure risk prediction index is in [0.00, 0.15], it is defined as a high-reliability nuclear grade range, and the judgment result is a qualified product for a first-level nuclear power key component. The system automatically associates this index with its unique production serial number, raw material furnace number, and visual inspection image sequence and stores it in the quality traceability database; when the failure risk prediction index is in (0.15, 0.65], it is defined as a controlled use range, and it is judged as a second-level non-core loop product, and the encrypted re-inspection process is triggered; when the failure risk prediction index is in (0.65, 1.00], it is defined as an immediate scrapping range. The system implements classification and interception through the production execution module, and at the same time as determining scrapping, it locks the corresponding production batch to start in-depth traceability of raw material composition and thermal processing technology.

[0061] Furthermore, the comprehensive reliability coefficient and the failure risk prediction index are mutually offset components, and their mathematical mapping logic is: the comprehensive reliability coefficient equals the difference between 1 and the failure risk prediction index. When the system detects an increase in the failure risk prediction index, the comprehensive reliability coefficient decreases synchronously, thereby triggering the interception mechanism in S5. Subsequent simulation experiments uniformly use the failure risk prediction index as the characterization object, and its evaluation conclusions are completely equivalent to the judgment criteria of the comprehensive reliability coefficient.

[0062] The specific calculation logic of the failure risk prediction index is described as follows: First, the extracted stress concentration coefficient is multiplied by the material's radiation embrittlement index to obtain the basic risk component; then, the difference between 1 and the probability of internal microstructure homogeneity is used as the exponential term, and the exponential function value is calculated with the natural constant as the base to obtain the microstructure deterioration weight; finally, the basic risk component and the microstructure deterioration weight are multiplied to obtain the final failure risk prediction index. This calculation logic ensures that when the microstructure homogeneity falls below the critical threshold, the risk index can experience a non-linear and rapid increase.

[0063] To verify the effectiveness of the detection logic proposed in this invention in identifying the correlation between "internal defects and external manifestations" in nuclear power plant gaskets, this embodiment constructs a virtual detection simulation platform based on multi-physics coupling. This platform simulates gasket samples under different processing thermal stress backgrounds and compares the output differences between traditional visual inspection methods and the method of this invention; please refer to Table 1 for details.

[0064] Table 1: Comparison of Simulation Experiment Data for Quality Inspection of Nuclear Power Plant Sealing Gaskets

[0065] Experimental group number Texture distortion rate Microstructure uniformity (probability value) Extreme working condition strength Tissue degradation coupling factors Overall reliability coefficient Traditional methods are used to determine the result (pass / fail). Experiment 1-1 0.012 0.98 15 / 300 0.0002 0.985 qualified Experiment 1-2 0.011 0.97 15 / 300 0.0003 0.982 qualified Experiment 2-1 0.045 0.65 15 / 300 0.0157 0.612 Pass (Misjudged) Experiment 2-2 0.042 0.62 15 / 300 0.0160 0.598 Pass (Misjudged) Experiment 3-1 0.125 0.45 15 / 300 0.0687 0.325 Unqualified Experiment 3-2 0.130 0.42 15 / 300 0.0754 0.298 Unqualified

[0066] In Table 1, the microstructure degradation coupling factor Ω is obtained by calculating the product of "texture distortion rate" and (1 - "microstructure uniformity probability"); this calculation process is performed within the nonlinear regression function in step S4; it is used to quantify the synergistic destructive capability of gasket surface process defects and internal physical state. The higher the value, the higher the risk of stress corrosion cracking of the gasket under high temperature and high pressure environment.

[0067] Table 1 compares the detection accuracy of Experiments 2-1 and 2-2. The texture distortion rate of both groups of samples is less than 0.05. Under the logic of traditional visual inspection methods, surface defects would not reach the trigger threshold and would be mistakenly judged as "qualified". However, this invention, through a mechanistic model, detects that the internal tissue uniformity is only about 0.65, and the calculated comprehensive reliability coefficients are 0.612 and 0.598, respectively. Referring to safety thresholds, such products belong to the high-risk category. This proves that this invention improves the ability to identify latent defects by approximately 38.5% compared to existing technologies. The calculation basis is: (Vth − Cinvention) / Vth, where Vth is the false negative threshold of traditional methods.

[0068] Data from Experiment 3 showed that when surface defects and poor internal microstructure uniformity coexist, the microstructure degradation coupling factor Ω increases exponentially. This confirms that the present invention is not merely a simple superposition of parameters, but rather achieves effective prediction of performance degradation under extreme environments through a multi-dimensional evaluation matrix.

[0069] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A machine vision-based method for quality inspection of nuclear power plant sealing gaskets, characterized in that, Includes the following steps: S1. Use a multispectral imaging device to acquire macroscopic surface image data of the nuclear power plant sealing gasket to be tested, and simultaneously establish a high-precision surface morphology mapping matrix to realize the digital characterization of the gasket surface geometry and minor defects. S2. Using a surface stress visual analysis algorithm, extract the texture distortion rate and edge gradient consistency parameters from the macroscopic surface image data to quantitatively characterize the processing consistency of the gasket surface and generate surface quality evaluation indicators. S3. Using a preset material physics mechanism model, the surface morphology mapping matrix is ​​correlated with the historical data of heat treatment in the production process to inversely calculate the probability of uniformity of microstructure inside the material and generate internal defect prediction index. S4. The surface quality evaluation index and the internal defect prediction index are weighted and fused to construct a multi-dimensional quality assessment matrix, and the comprehensive reliability coefficient is calculated. S5. Based on the comparison result between the comprehensive reliability coefficient and the preset safety threshold, the gasket quality level is automatically output and fed back to the production execution for classification and interception or process parameter compensation.

2. The method for quality inspection of nuclear power plant sealing gaskets based on machine vision according to claim 1, characterized in that: In step S1, a multi-axis industrial camera deployed above the production line collects reflected light intensity data from multiple angle light sources, and structured light scanning technology is used to obtain spatial depth information of each coordinate point on the gasket surface, thereby generating a high-precision surface topography mapping matrix.

3. The method for quality inspection of nuclear power plant sealing gaskets based on machine vision according to claim 1, characterized in that: In step S2, the regularity of surface texture is analyzed by calculating the gray-level co-occurrence matrix between adjacent pixels in the image, and micron-level cracks and indentations are identified using an edge detection operator. The area, depth, and distribution density of the identified defects are used as variables, and after dimensionless processing, surface quality evaluation indicators are obtained.

4. The method for quality inspection of nuclear power plant sealing gaskets based on machine vision according to claim 1, characterized in that: In S3, the material physical mechanism model is based on the thermoelastic mechanics equation. It extracts the micro-undulation features in the surface morphology as the cause of thermal stress concentration, and combines the historical data of compositional segregation of this batch of metal materials to assess the risk of internal lattice defects caused by thermal stress.

5. The method for quality inspection of nuclear power plant sealing gaskets based on machine vision according to claim 1, characterized in that: In S4, the calculation logic of the comprehensive reliability coefficient is as follows: the surface quality evaluation index is defined as the explicit risk component, the internal defect prediction index is defined as the implicit failure component, the coupling correlation value between the two is calculated through a nonlinear regression function, and the performance degradation correction factor under extreme working conditions is introduced to finally obtain the comprehensive reliability coefficient reflecting the long-term sealing capability.

6. The method for quality inspection of nuclear power plant sealing gaskets based on machine vision according to claim 1, characterized in that: In S5, the criteria for automatically outputting the quality level of the gasket are as follows: when the comprehensive reliability coefficient is in the first range, it is marked as a nuclear-grade qualified product; when only the surface indicators are qualified but the internal predicted indicators are abnormal, it is marked as a downgraded product or a product subject to mandatory re-inspection; when there are obvious defects on the surface and the internal risk is high, it is marked as a scrapped product.

7. The method for quality inspection of nuclear power plant sealing gaskets based on machine vision according to claim 3, characterized in that: The process of generating the surface quality evaluation index also includes: performing a Fourier transform on the collected light intensity distribution, extracting high-frequency components in the frequency domain to identify surface micro-irregularities, and comparing them with the spectral characteristics of standard samples to eliminate interference from human experience under different testing environments.

8. A machine vision-based method for quality inspection of nuclear power plant sealing gaskets according to claim 4, characterized in that: The calculation logic of the internal defect prediction index is as follows: establish the mapping relationship between surface geometrical abrupt changes and internal stress field distribution, calculate the internal stress concentration factor caused by surface micro-deformation, and multiply the factor by the radiation embrittlement index of the material to obtain the prediction index characterizing the failure risk of the gasket in a strong radiation environment.

9. A machine vision-based method for quality inspection of nuclear power plant sealing gaskets according to claim 1, characterized in that: The S5 also includes establishing a closed-loop quality traceability logic, which associates and stores the comprehensive reliability coefficient of each gasket with its unique production serial number, raw material furnace number and visual inspection image sequence to ensure the traceability of the data source.

10. A machine vision-based method for quality inspection of nuclear power plant sealing gaskets according to claim 5, characterized in that: The calculation of the coupling correlation value is achieved by analyzing the superposition effect of surface process defects and internal component segregation under high temperature and high pressure. If the internal structure uniformity corresponding to the area where the surface microcrack is located is lower than the preset uniformity threshold, the deduction weight of the comprehensive reliability coefficient is increased exponentially.

Images