Intelligent quality measurement method and system for a product manufacturing process

Abstract

The present application belongs to the technical field of intelligent quality measurement, and specifically relates to an intelligent quality measurement method and system for a product manufacturing process. The method comprises the following steps: obtaining marked risk distribution data by processing a preliminary distribution matrix of surface material components; obtaining an original waveform set of internal stress distribution by extracting the spatial position of a high-risk area from the marked risk distribution data; obtaining a preliminary grid map of stress distribution according to the original waveform set of internal stress distribution; obtaining an internal stress non-uniformity distribution map according to the preliminary grid map of stress distribution; obtaining comprehensive quality risk distribution data by aligning the spatial positions of the preliminary distribution matrix of surface material components and the internal stress non-uniformity distribution map; and obtaining an optimized processing control sequence by comprehensively judging the comprehensive quality risk distribution data. Compared with a traditional offline sampling inspection mode, the method effectively solves the problems of one-sidedness, lag and inability to analyze in a linked manner in the traditional detection.

Description

Technical Field

[0001] This invention belongs to the field of intelligent quality measurement technology, specifically an intelligent quality measurement method and system for product manufacturing process. Background Technology

[0002] In modern manufacturing, precise control of product quality is a key pillar for enhancing corporate competitiveness and market trust. With the continuous improvement of industrial intelligence, real-time monitoring and assurance of product quality during production has become an important direction for the transformation and upgrading of the manufacturing industry. Especially in high-precision, high-requirement manufacturing fields, quality measurement is not only the standard for determining product qualification, but also a core element for optimizing processes and reducing costs. However, traditional quality inspection methods often face numerous challenges and urgently require innovative breakthroughs.

[0003] Current quality measurement methods largely rely on offline sampling or single physical tests. While these methods can identify problems to some extent, they are ill-suited to the changing product characteristics and dynamic process conditions in complex production environments. Especially when dealing with different materials and processing stages, existing methods often fail to fully capture subtle changes in the product during manufacturing, leading to overlooked quality risks and ultimately affecting the reliability and consistency of the finished product. This limitation makes a new technology that can comprehensively and in real-time sense the product's status on the production line urgently needed in the manufacturing industry.

[0004] Against this backdrop, quality measurement during the manufacturing process faces significant technical challenges. A core issue is the dynamic change in the surface material composition of a product, as this directly determines its performance. During processing, materials may undergo unpredictable changes due to conditions such as heat and pressure. Furthermore, these compositional changes can lead to uneven internal stress distribution, affecting key indicators such as hardness and durability. For example, in the machining of metal parts, when the surface material composition undergoes subtle changes due to high temperatures, internal stress may concentrate in certain areas, causing cracks or deformation in the parts during subsequent use. Such problems are often difficult to detect in a timely manner during production.

[0005] Therefore, how to capture changes in the material composition of the product surface in real time during the manufacturing process and accurately perceive the non-uniformity of internal stress distribution has become a key issue in improving the intelligence of quality measurement. Summary of the Invention

[0006] To address the above problems, this invention provides an intelligent quality measurement method and system for the product manufacturing process, which solves the problem of how to capture changes in the composition of the product surface material in real time during the manufacturing process and accurately perceive the non-uniformity of the internal stress distribution.

[0007] To achieve the above objectives, the technical solution adopted by the present invention is as follows: In a first aspect, the present invention provides an intelligent quality measurement method for a product manufacturing process, comprising: By continuously scanning the product surface with a spectrometer, a preliminary distribution matrix of the surface material composition is obtained; By processing the preliminary distribution matrix of the surface material composition, the labeled risk distribution data is obtained; The spatial location of high-risk areas is extracted from the labeled risk distribution data to obtain the original waveform set of internal stress distribution; Based on the original waveform set of the internal stress distribution, a preliminary mesh diagram of the stress distribution is obtained; Based on the preliminary grid diagram of stress distribution, the internal stress non-uniformity distribution diagram is obtained; By aligning the preliminary distribution matrix of surface material composition with the internal stress non-uniformity distribution map in terms of spatial location, comprehensive quality risk distribution data is obtained. Based on the comprehensive quality risk distribution data, a comprehensive judgment is made to obtain the optimized processing control sequence.

[0008] The preferred technical solution of this application has the following advantages: A product surface composition distribution matrix is ​​obtained through spectral scanning; after risk marking, high-risk area extraction, stress waveform assembly construction, and stress mesh generation, an internal stress non-uniformity distribution map is obtained; then, the composition distribution and stress distribution are spatially aligned and fused to form a comprehensive quality risk distribution and output an optimized processing control sequence. Compared to traditional offline sampling inspection methods, this approach can comprehensively and accurately capture changes in material composition and internal stress online, promptly identify potential quality issues, significantly improve product reliability and production consistency, and effectively solve the problems of limited, delayed, and incompatible analysis in traditional testing methods.

[0009] As a preferred embodiment of the intelligent quality measurement method for a product manufacturing process according to the present invention, wherein: the obtained labeled risk distribution data includes: An initial spatiotemporal feature mapping matrix is ​​constructed to obtain preliminary spectral signal distribution data; Based on the preliminary spectral signal distribution data, the fluctuation amplitude data of the signal at different time points and spatial locations are determined; If the fluctuation amplitude data exceeds the preset fluctuation amplitude threshold range, then obtain the distribution data of the marked abnormal areas; Based on the marked abnormal area distribution data, determine whether the abnormal area belongs to the high-risk category to obtain the classified risk area data; The specific location information of high-risk areas is extracted from the classified risk area data. By comparing it with the spatiotemporal feature mapping matrix, the corresponding range of high-risk areas in the dynamic matrix is ​​determined, and the spatiotemporal distribution data of high-risk areas is obtained. In-depth analysis of the peak characteristics of the spatiotemporal distribution data spectral signals of high-risk areas is conducted to obtain peak characteristic distribution data; By using peak feature distribution data, high-risk areas are finally labeled to generate complete risk distribution data.

[0010] The preferred technical solution of this application has the following advantages: by constructing a spatiotemporal feature mapping matrix and combining it with signal fluctuation amplitude to identify abnormal areas, high-risk areas are classified and located, and spectral peak characteristics are analyzed in depth to achieve accurate risk labeling. It can capture dynamic changes in material composition in real time, accurately locate potential hazard areas, effectively solve the problems of traditional detection lag and inability to identify subtle component anomalies and risk areas in real time, and improve product quality stability.

[0011] As a preferred embodiment of the intelligent quality measurement method for a product manufacturing process according to the present invention, wherein: obtaining the original waveform set of internal stress distribution includes: By extracting the spatial location of high-risk areas from risk distribution data, a preliminary echo data set is obtained; Based on the acquired preliminary echo data set, data cleaning is performed to obtain a cleaned waveform signal group; For the cleaned waveform signal group, determine the corrected waveform dataset; Preliminary mapping data of stress distribution are obtained from the corrected waveform dataset; By using preliminary mapping data of stress distribution, potential areas of high stress concentration can be identified; If the classification results show that the high stress concentration area exceeds the preset stress threshold, then the original waveform data of the current area will be analyzed a second time. Based on the stress distribution characteristics obtained from the secondary analysis, a detailed stress distribution map of the corresponding high-risk area is generated, the final detection result is determined, and the original waveform set of the internal stress distribution is generated.

[0012] As a preferred embodiment of the intelligent quality measurement method for the product manufacturing process described in this invention, wherein obtaining the preliminary mesh diagram of stress distribution includes: Based on the waveform data related to internal stress, simulation results of the propagation path are obtained; Based on the simulation results of the propagation path, spatial units are divided to determine the potential regions of stress anomalies; For potential areas of stress anomalies, obtain the specific coordinate information of the concentration points; By using the specific coordinate information of the concentration points and the results of spatial unit division, a preliminary grid diagram of stress distribution is constructed, and the distribution pattern of stress anomalies in the grid diagram is determined. Based on the distribution pattern of stress anomalies in the mesh diagram, a preliminary mesh diagram of stress distribution is generated.

[0013] As a preferred embodiment of the intelligent quality measurement method for the product manufacturing process described in this invention, the step of determining the distribution pattern of stress anomalies in the mesh diagram includes: If the stress anomaly distribution in the mesh diagram shows a concentrated trend, then the spatial units around the concentrated points are refined into a finer mesh to obtain the stress anomaly boundary and thus the refined mesh distribution data. Based on the refined grid distribution data, the support vector machine algorithm is used to classify the abnormal areas, distinguish different levels of stress concentration, and determine the stress distribution map. Analyze the correlation between stress anomalies and concentration points within spatial cells in the stress distribution map to generate detailed mapping results of stress distribution.

[0014] As a preferred embodiment of the intelligent quality measurement method for the product manufacturing process described in this invention, the step of obtaining the internal stress non-uniformity distribution map includes: By extracting stress distribution data from the preliminary grid diagram of stress distribution, the characteristics of local stress distribution are analyzed. Based on the characteristics of local stress distribution, the non-uniform distribution results are calculated and determined; For the results of non-uniform distribution, the distribution density of stress concentration points is analyzed. If the distribution density is higher than a preset threshold, the subsequent visualization processing is triggered to obtain the density determination result. Data from high-density areas are extracted from the density determination results to generate preliminary stress visualization graphics; Obtain preliminary stress visualization graphics and generate the final internal stress non-uniformity distribution diagram; By analyzing the relationship between local and overall distribution through the final internal stress non-uniformity distribution diagram, cluster analysis is used to classify the non-uniform regions and obtain the classification distribution results. Based on the classification and distribution results, a targeted analysis file is generated to record the characteristic data of each type of non-uniform region, thus completing a comprehensive analysis of the stress distribution.

[0015] As a preferred embodiment of the intelligent quality measurement method for a product manufacturing process described in this invention, the step of triggering the subsequent visualization processing includes: A heatmap algorithm is used to color-map the stress values ​​of each mesh element according to the color gradient; An interpolation algorithm is introduced to smooth the color distribution and generate a heat map of the basic stress distribution; By combining the gradient matrix and stress concentration point information, layers are overlaid on the heat map; Regions with local gradients greater than 30 MPa are highlighted with black borders to obtain the final internal stress non-uniformity distribution map.

[0016] As a preferred embodiment of the intelligent quality measurement method for a product manufacturing process described in this invention, the step of obtaining comprehensive quality risk distribution data includes: By reading the preliminary distribution matrix of surface materials and their components, spatial location information in the initial matrix is ​​obtained, and the component data corresponding to each location point is determined.

[0017] Based on the obtained spatial location information, the internal stress is matched with the non-uniform distribution pattern to obtain stress distribution data corresponding to the initial matrix; For high-risk areas, the corresponding component data and stress distribution data are extracted, and the correlation value between the two in spatial location is calculated. Based on the correlation value, the comprehensive quality risk distribution data is determined.

[0018] As a preferred embodiment of the intelligent quality measurement method for a product manufacturing process described in this invention, the step of determining the comprehensive quality risk distribution data based on correlation values ​​includes: If the correlation value exceeds the preset threshold range, a local analysis of the gradient changes in the high-risk area is performed to determine whether there are abnormal fluctuation areas. By analyzing the correlation between abnormal fluctuation areas and quality risks, the risk level is classified using the support vector machine algorithm, and the risk distribution results for each area are obtained. Based on the classification results, a corresponding quality risk distribution view is generated for the gradient change data of the marked locations in high-risk areas, and the final risk assessment data is determined. If outliers are found in the risk assessment data, a secondary verification is performed by combining the spatial location and the non-uniform distribution graph to obtain corrected quality risk distribution data.

[0019] Secondly, the present invention provides an intelligent quality measurement system for a product manufacturing process, comprising: The spectral scanning module continuously scans the product surface using a spectral analyzer to obtain a preliminary distribution matrix of the surface material composition. The composition labeling module processes the preliminary distribution matrix of the surface material composition to obtain labeled risk distribution data; The waveform extraction module extracts the spatial location of high-risk areas from the marked risk distribution data to obtain the original waveform set of internal stress distribution. The stress mesh module generates a preliminary mesh diagram of the stress distribution based on the original waveform set of the internal stress distribution. The non-uniformity analysis module obtains an internal stress non-uniformity distribution map based on the preliminary mesh map of stress distribution; The spatial alignment module aligns the preliminary distribution matrix of surface material composition with the internal stress non-uniformity distribution map in spatial position to obtain comprehensive quality risk distribution data. The processing optimization module makes a comprehensive judgment based on the overall quality risk distribution data to obtain an optimized processing control sequence.

[0020] The beneficial effects of an intelligent quality measurement system for product manufacturing processes according to an embodiment of this application are as follows: Through multi-module collaboration, it achieves visual intelligent quality monitoring throughout the entire product manufacturing process, effectively solving the problems of traditional offline sampling inspection, single inspection dimensions, and inability to capture material composition and internal stress changes in real time. The spectral scanning module can continuously collect surface spectral information of the product online, quickly generating a material composition distribution matrix, replacing traditional offline sampling methods, and realizing dynamic monitoring during the production process. The composition marking module marks the composition distribution data for risks, accurately identifying abnormal areas and avoiding the omission of quality hazards caused by subtle fluctuations in composition. The waveform extraction module and stress grid module can invert the original waveform of internal stress from the risk areas and construct a stress grid, intuitively reflecting the internal stress distribution state; the non-uniformity analysis module further quantifies the degree of stress non-uniformity, making up for the deficiency of traditional detection in associating composition and stress. The spatial alignment module accurately matches surface composition data and stress non-uniformity data in space, forming a comprehensive quality risk distribution, making quality judgment more consistent with the actual state of the product. The processing optimization module outputs an optimized processing control sequence based on the comprehensive risk results, realizing closed-loop control from monitoring and analysis to process adjustment. The overall system can perceive subtle changes in the product during processing in real time and comprehensively, promptly detect potential risks such as cracks and deformation, significantly improve product reliability and consistency, and meet the needs of high-precision manufacturing for intelligent and online quality control. Attached Figure Description

[0021] Figure 1 A flowchart illustrating an intelligent quality measurement method for a product manufacturing process; Figure 2 This is a flowchart illustrating an intelligent quality measurement system for a product manufacturing process. Detailed Implementation

[0022] To enable those skilled in the art to better understand the technical solution, the present invention will be described in detail below with reference to embodiments. The description in this part is only exemplary and explanatory, and should not be used to limit the scope of protection of the present invention in any way.

[0023] Example 1, as Figure 1 As shown in the illustration, an embodiment of the present invention provides an intelligent quality measurement method for a product manufacturing process, comprising: S1: The surface of the product is continuously scanned by a spectrometer to obtain a preliminary distribution matrix of the surface material composition; S2: By processing the preliminary distribution matrix of the surface material composition, the labeled risk distribution data is obtained; S3: Extract the spatial location of high-risk areas from the labeled risk distribution data to obtain the original waveform set of internal stress distribution; S4: Based on the original waveform set of the internal stress distribution, a preliminary mesh diagram of the stress distribution is obtained; S5: Based on the preliminary grid diagram of stress distribution, obtain the internal stress non-uniformity distribution diagram; S6: By spatially aligning the preliminary distribution matrix of surface material composition with the internal stress non-uniformity distribution map, comprehensive quality risk distribution data is obtained; S7: Based on the comprehensive quality risk distribution data, a comprehensive judgment is made to obtain the optimized processing control sequence.

[0024] It should be noted that, in the embodiments of this application, the product surface is continuously scanned online using a spectrometer in step S1 to directly obtain the preliminary distribution matrix of the surface material composition. This enables real-time perception of the dynamic changes in material composition, replacing the traditional lagging and one-sided sampling inspection mode. It can capture changes in material composition caused by factors such as heat and pressure during the production process in a timely manner, providing reliable raw data support for subsequent quality analysis.

[0025] In step S2, the composition distribution matrix is ​​processed and risk distribution data is labeled, enabling precise identification of abnormal signals and suspicious areas. This solves the problem that traditional detection methods struggle to quantify and determine the degree of material anomalies, achieving a shift from passive detection to proactive early warning. Steps S3 to S5 further extract the original waveforms of internal stress based on high-risk areas, construct a stress mesh map, and generate an internal stress non-uniformity distribution map. This correlates surface composition changes with internal stress states, overcoming the technical bottleneck of traditional methods that cannot simultaneously monitor material composition and internal stress. It enables early identification of risks such as cracks and deformations that may be caused by stress concentration.

[0026] Step S6 integrates surface material composition distribution and internal stress non-uniformity information through spatial alignment to form comprehensive quality risk distribution data. This enables collaborative analysis of multi-dimensional quality information, making risk assessment more comprehensive and accurate. Finally, in step S7, intelligent decision-making is performed based on the comprehensive risk data, outputting an optimized processing control sequence to form a closed-loop control mechanism of monitoring-analysis-decision-optimization.

[0027] In summary, the embodiments of this application can identify potential quality problems in the product manufacturing process in real time, comprehensively and accurately, effectively improve product reliability and consistency, reduce defect rate, meet the intelligent quality control needs of high-precision manufacturing, and promote the upgrading of the manufacturing quality monitoring mode from offline and post-event detection to online, real-time and intelligent management and control.

[0028] Example 2, as Figure 1 As shown in the illustration, an embodiment of the present invention provides an intelligent quality measurement method for a product manufacturing process, comprising: S1: The surface of the product is continuously scanned by a spectrometer to obtain a preliminary distribution matrix of the surface material composition.

[0029] It should be noted that the product surface is continuously scanned by a spectrometer, and the reflectance spectral signal is collected for dynamically changing areas. The signal is segmented into different wavelength ranges using spectral band division technology. Background noise is removed and normalized by reflection intensity calibration to obtain a preliminary distribution matrix of surface material composition.

[0030] Furthermore, a continuous scanning operation is performed on the product surface using a spectrometer to collect reflectance spectral signals in dynamically changing areas, constructing an initial spectral data set. Spectral band segmentation technology is used to segment the collected reflectance spectral signals according to wavelength ranges, generating segmented spectral signal groups. For these segmented spectral signal groups, a reflectance intensity calibration method is applied to remove background noise, resulting in a calibrated spectral signal set. The calibrated spectral signal set is then normalized to eliminate signal intensity differences, constructing a standardized spectral data matrix. Based on the standardized spectral data matrix and a preset threshold, signals are classified. If the signal intensity of a certain band exceeds the threshold, it is classified as a significant feature signal, determining the preliminary distribution matrix of the surface material composition. Spatial mapping processing is performed on the preliminary distribution matrix to generate a spatial distribution map of the surface material composition, obtaining detailed compositional distribution information for dynamically changing areas. If there are missing signal regions in the spatial distribution map, a pre-established interpolation model is used to fill in the missing data, obtaining a complete surface material composition distribution result.

[0031] It should be noted that when continuously scanning the product surface using a spectrometer, a high-resolution spectrometer can be used, moving across the surface at a scanning frequency of 100 times per second, covering an area of ​​50 square centimeters. Reflectance spectral signals with wavelengths ranging from 400 nanometers to 2500 nanometers are collected. For dynamically changing areas, image processing algorithms are used in conjunction with the spectral signal change rate, setting a change rate threshold of 0.05 to filter out areas with large signal fluctuations as key analysis targets. Next, spectral band segmentation technology is employed to divide the 400-2500 nanometer range into five sub-bands, such as 400-700 nanometers for the visible light band and 700-1100 nanometers for the near-infrared band. Fourier transform is used to perform frequency domain analysis on the signal in each band, extracting characteristic frequencies to segment the signal and ensuring that signal characteristics from different wavelength ranges are not confused. Subsequently, reflection intensity calibration is performed. Using preset background spectral data (assuming a mean background noise intensity of 0.1 and a standard deviation of 0.02), background noise is removed through subtraction, and the signal intensity is adjusted to the range of 0 to 1 using a min-max normalization method. For example, if the original signal intensity is 0.3, the normalized intensity is 0.6, ensuring data consistency. Finally, based on the normalized signal data, principal component analysis (PCA) is used to extract the first three principal components, assuming contribution rates of 60%, 25%, and 10%, respectively, to construct a preliminary distribution matrix of the surface material composition. The matrix has a dimension of 50×50, corresponding to each pixel in the scanned area. By calculating the principal component score for each point, the material composition distribution is preliminarily determined. For example, areas with scores higher than 0.8 may be metallic, while those lower may be non-metallic. To establish a logical chain, the distribution matrix can be compared with the product design database to verify whether the material distribution meets expectations. If anomalies are found (accounting for more than 5%), subsequent depth spectral analysis is triggered to ensure the reliability of the analysis results.

[0032] S2: By processing the preliminary distribution matrix of the surface material composition, the labeled risk distribution data is obtained.

[0033] It should be noted that, based on the preliminary distribution matrix of the surface material composition, a spatiotemporal feature mapping is constructed to generate a dynamic matrix of spectral signals that vary with time and space. Significant absorption peaks or reflection peaks are identified by peak feature extraction. If the signal fluctuation amplitude in the dynamic matrix exceeds a preset dynamic change threshold, the abnormal area is marked as a high-risk area, and the marked risk distribution data is obtained.

[0034] In this embodiment of the application, obtaining the labeled risk distribution data in step S2 includes steps A1-A7: A1: Construct the initial spatiotemporal feature mapping matrix to obtain preliminary spectral signal distribution data; A2: Based on the preliminary spectral signal distribution data, determine the fluctuation amplitude data of the signal at different time points and spatial locations; A3: If the fluctuation amplitude data exceeds the preset fluctuation amplitude threshold range, then obtain the distribution data of the marked abnormal areas; A4: Based on the marked abnormal area distribution data, determine whether the abnormal area belongs to the high-risk category to obtain the classified risk area data; A5: Extract the specific location information of high-risk areas from the classified risk area data, and determine the corresponding range of high-risk areas in the dynamic matrix by comparing it with the spatiotemporal feature mapping matrix, thereby obtaining the spatiotemporal distribution data of high-risk areas; A6: Conduct in-depth analysis of the peak characteristics of the spatiotemporal distribution data spectral signals of high-risk areas to obtain peak characteristic distribution data; A7: By using peak feature distribution data, high-risk areas are finally labeled to generate complete risk distribution data.

[0035] Furthermore, the specific implementation methods for steps A1-A7 are as follows: By processing the compositional distribution data of the surface material, an initial spatiotemporal feature mapping matrix is ​​constructed. Signal decomposition technology is used to separate the main components in the spectral signal, obtaining preliminary spectral signal distribution data. Based on this preliminary spectral signal distribution data, a dynamic matrix varying with time and space is generated. Point-by-point analysis of the signal change trend in the dynamic matrix is ​​performed to determine the fluctuation amplitude data of the signal at different time points and spatial locations. If the fluctuation amplitude data exceeds a preset fluctuation amplitude threshold, the corresponding time point and spatial location are marked as anomalies, obtaining the distribution data of the marked anomaly regions. For the marked anomaly region distribution data, combined with the compositional distribution characteristics of the surface material, a support vector machine algorithm is used to classify the anomaly regions, determining whether they belong to the high-risk category, obtaining the classified risk region data. The specific location information of high-risk regions is extracted from the classified risk region data. By comparing it with the spatiotemporal feature mapping matrix, the corresponding range of the high-risk regions in the dynamic matrix is ​​determined, obtaining the spatiotemporal distribution data of the high-risk regions. After obtaining the spatiotemporal distribution data of the high-risk regions, in-depth analysis is performed on the peak characteristics of their spectral signals to identify significant absorption peaks and reflection peaks, obtaining peak feature distribution data. By combining peak feature distribution data with risk labeling information, high-risk areas are finally labeled to generate complete risk distribution data.

[0036] Furthermore, regarding the preliminary distribution matrix of the surface material composition, we first assume a 10×10 two-dimensional matrix representing the spatial distribution, where each element value represents the concentration of a material component at a specific location, ranging from 0 to 100. For example, a value of 65.5 indicates a high concentration of a certain component at that point. Next, we construct a spatiotemporal feature mapping, introducing the time dimension. Assuming a time step of 1 second, we collect data at 10 time points, forming a 10×10×10 three-dimensional dynamic matrix. The matrix element values ​​change over time; for example, at a certain location, the concentration value changes from 65.5 to 70.2 at the 5th second, reflecting the dynamic trend. When generating the spectral signal, we use a Fourier transform algorithm to perform frequency domain analysis on the time series of each spatial point, extracting signal features in the frequency range of 0.1 to 5 Hz, and calculating the spectral intensity of each point. Assuming a point has a dominant frequency of 2.3 Hz and an intensity of 8.7, it is recorded as a significant feature. Subsequently, peak feature extraction is performed, and the signal is smoothed using Gaussian filtering. A peak threshold of 5.0 is set to identify absorption or reflection peaks. For example, if the spectral intensity at a certain point is 8.7, which is greater than the threshold, it is marked as a significant absorption peak. If the signal fluctuation amplitude data in the dynamic matrix exceeds the preset fluctuation amplitude threshold range, for example, if the threshold is set to 10%, and the concentration value of a certain area suddenly increases from 65.5 to 80.0 at a certain time point, with a fluctuation rate of 22.1%, it is judged as an anomaly. This area is marked as high-risk, and risk distribution data is generated. Assuming the coordinates of this area are (3,4), the mark value is 1, indicating high risk, while other areas are marked as 0, forming the final risk distribution matrix. By forming a complete logical chain from data acquisition to feature extraction to risk assessment, each link is automatically processed based on numerical values ​​and algorithms. The marking of abnormal areas can also be linked to subsequent material defect detection operations to further optimize risk management.

[0037] S3: Extract the spatial location of high-risk areas from the labeled risk distribution data to obtain the original waveform set of internal stress distribution.

[0038] It should be noted that the spatial location of high-risk areas is extracted from the marked risk distribution data. An ultrasonic sensor is used to emit sound waves and receive echo signals within the corresponding area. The echo signals are filtered to remove high-frequency interference through ultrasonic waveform preprocessing. The time delay and amplitude are adjusted in combination with the echo signals to obtain the original waveform set of internal stress distribution.

[0039] In this embodiment of the application, obtaining the original waveform set of internal stress distribution in step S3 includes steps B1-B7: B1: By extracting the spatial location of high-risk areas from risk distribution data, a preliminary echo data set is obtained; B2: Based on the acquired preliminary echo data set, perform data cleaning to obtain a cleaned waveform signal group; B3: For the cleaned waveform signal group, determine the corrected waveform dataset; B4: Obtain preliminary mapping data of stress distribution from the corrected waveform dataset; B5: Identify potential high-stress concentration areas using preliminary stress distribution mapping data; B6: If the classification results show that the high stress concentration area exceeds the preset stress threshold, then perform a secondary analysis on the original waveform data of the current area; B7: Based on the stress distribution characteristics obtained from the secondary analysis, generate detailed stress distribution maps of the corresponding high-risk areas, determine the final detection results, and generate the original waveform set of internal stress distribution.

[0040] Furthermore, the specific implementation methods for steps B1-B7 are as follows: By extracting the spatial locations of high-risk areas from risk distribution data, an ultrasonic sensor is used to emit and receive echo signals in designated areas, obtaining a preliminary echo data set. Based on the acquired echo data set, filtering techniques are used to preprocess the echo signals, removing high-frequency interference to obtain a cleaned waveform signal set. For the cleaned waveform signal set, time delay and amplitude correction methods are combined to adjust the signal parameters, determining the corrected waveform dataset. From the corrected waveform dataset, signal features related to internal stress are extracted to obtain preliminary stress distribution mapping data. Using the preliminary mapping data, a support vector machine algorithm is used to classify the distribution patterns of internal stress and identify potential high-stress concentration areas. If the classification results show that high-stress concentration areas exceed a preset threshold, a secondary analysis of the original waveform data for that area is performed to obtain more refined stress distribution features. Based on the stress distribution features obtained from the secondary analysis, a detailed stress distribution map of the corresponding high-risk area is generated, determining the final detection result.

[0041] Furthermore, when extracting the spatial location of high-risk areas from the labeled risk distribution data, spatial analysis can be performed on the data using Geographic Information System (GIS) software. Assuming the risk distribution data is stored in a grid format, with each grid cell measuring 1 meter × 1 meter, and the risk value ranging from 0 to 100, with a threshold of 80, grid cells with a risk value greater than 80 are marked as high-risk areas. The coordinates of their center points (e.g., latitude and longitude data, accurate to 0.0001 degrees) are extracted, and the boundary polygons of the high-risk areas are generated for subsequent positioning. Next, within the corresponding area, ultrasonic sensors emit sound waves and receive echo signals. Assuming the sensor operates at a frequency of 40 kHz, the emitted sound wave power is 10 mW, and the coverage radius is 5 meters, the area is scanned using a sensor array (one sensor every 0.5 meters). The reception time and intensity of the echo signals are recorded, with a time resolution of 1 microsecond and an intensity range of 0 to 5V. Subsequently, the echo signal was preprocessed and filtered to remove high-frequency interference. A Butterworth low-pass filter with a cutoff frequency of 50kHz and a filter order of 4 was used. The signal was converted from the time domain to the frequency domain using a fast Fourier transform to filter out frequency components higher than 50kHz. Then, it was inversely transformed back to the time domain to obtain a smooth signal. The analysis results showed that the signal-to-noise ratio was improved by about 15% after filtering. Next, the time delay and amplitude were adjusted in conjunction with the echo signal correction. For the time delay between the sensor arrays, the cross-correlation algorithm was used to calculate the time difference between the signals of each sensor. Assuming a maximum time difference of 20 microseconds, the signal was time-aligned using linear interpolation. For amplitude correction, based on the sensor distance attenuation model (amplitude decreases inversely with the square of the distance), the signal amplitude at a distance of 5 meters was corrected to the standard value of 1V. After correction, the signal amplitude error was controlled within ±0.05V. Finally, the original waveform set of internal stress distribution is obtained. Through waveform feature extraction algorithm, the peak value (assuming the peak value range is 0.8V to 2.5V) and propagation time (range of 50 microseconds to 200 microseconds) of each waveform are calculated. Combined with the speed of sound (set to 1500 m / s), the spatial location of stress distribution is estimated. The analysis shows that the overlap between the stress concentration area and the high-risk area boundary is more than 90%, forming a complete logical chain from risk identification to stress distribution calculation.

[0042] S4: Based on the original waveform set of the internal stress distribution, a preliminary mesh diagram of the stress distribution is obtained.

[0043] It should be noted that, for the original waveform set of internal stress distribution, sound wave propagation modeling is used to simulate the propagation path and attenuation of sound waves inside the material. By meshing the stress distribution, the interior of the material is divided into spatial grid units, the location coordinates of the concentration points of stress anomalies are determined, and a preliminary grid diagram of stress distribution is obtained.

[0044] In this embodiment of the application, obtaining the preliminary mesh diagram of stress distribution in step S4 includes steps C1-C5: C1: Simulation results of the propagation path are obtained from waveform data related to internal stress; C2: Based on the simulation results of the propagation path, divide the space into units to determine the potential areas of stress anomalies; C3: For potential areas of stress anomalies, obtain the specific coordinate information of the concentration points; C4: By using the specific coordinate information of the concentration points and combining the results of spatial unit division, a preliminary grid diagram of stress distribution is constructed, and the distribution pattern of stress anomalies in the grid diagram is determined. C5: Based on the distribution pattern of stress anomalies in the mesh diagram, generate a preliminary mesh diagram of stress distribution.

[0045] In this embodiment of the application, the distribution pattern of stress anomalies in the mesh diagram in step C4 is as follows: If the stress anomaly distribution in the mesh diagram shows a concentrated trend, then the spatial units around the concentrated points are refined into a finer mesh to obtain the stress anomaly boundary and thus the refined mesh distribution data. Based on the refined grid distribution data, the support vector machine algorithm is used to classify the abnormal areas, distinguish different levels of stress concentration, and determine the stress distribution map. Analyze the correlation between stress anomalies and concentration points within spatial cells in the stress distribution map to generate detailed mapping results of stress distribution.

[0046] Furthermore, the specific implementation methods for steps C1-C5 are as follows: For waveform data related to internal stress, a sound wave propagation model is used to simulate the propagation path and attenuation characteristics of sound waves within the material, calculating the intensity variation of sound waves at different locations to obtain simulation results of the propagation path. Based on the simulation results, the interior of the material is divided into multiple spatial units through meshing, and the distribution of sound wave intensity within each unit is analyzed to identify potential areas of stress anomalies. For these potential areas, waveform data features within each spatial unit are extracted to locate the concentration points of anomalies and obtain their specific coordinate information. Using the coordinate information of the concentration points, combined with the spatial unit division results, a preliminary mesh map of stress distribution is constructed to determine the distribution pattern of stress anomalies in the mesh map. If the stress anomaly distribution in the mesh map shows a concentrated trend, the mesh of the spatial units surrounding the concentration points is refined to obtain more accurate stress anomaly boundaries, resulting in refined mesh distribution data. Based on the refined mesh distribution data, a support vector machine algorithm is used to classify the anomaly regions, distinguishing different levels of stress concentration, and determining the final stress distribution map. By analyzing the correlation between stress anomalies and concentration points within the spatial unit using the final stress distribution map, detailed mapping results of the stress distribution are generated.

[0047] Furthermore, for the original waveform set of internal stress distribution, the propagation path and attenuation of sound waves inside the material are first simulated by sound wave propagation modeling. Specifically, a three-dimensional material model is established using finite element analysis software. The material is assumed to be homogeneous and isotropic. The sound wave frequency is set to 20kHz, the sound velocity is 5000m / s, and the initial amplitude is 1Pa. The propagation trajectory of the sound wave after it is incident from one side of the material surface is simulated. The energy attenuation of the sound wave due to material damping during propagation is calculated, and a distribution map with an attenuation coefficient of 0.05 / m is obtained. The propagation time and intensity changes of the sound wave at each point are recorded to form a propagation path dataset. Subsequently, the internal stress distribution of the material was meshed. Specifically, the material was divided into cubic spatial grid cells with a side length of 0.01m, generating a total of 1 million grid points. By analyzing outliers in the sound wave propagation time within each grid cell and combining this with sound wave intensity attenuation data, the stress value of each grid cell was calculated. Assuming that stress is directly proportional to the sound wave propagation time delay, with a 0.1ms increase in delay corresponding to a 10MPa increase in stress, a stress distribution matrix for each grid cell was obtained. Next, the coordinates of the stress anomaly concentration points were determined. Using a clustering algorithm such as K-means, grid cells with stress values ​​1.5 times higher than the average were classified as outliers. The number of cluster centers was set to 5, and the coordinates of the concentration points were iteratively calculated. For example, a concentration point located at (0.25m, 0.3m, 0.15m) had a stress peak of 150MPa. The stress gradient changes of the surrounding grid cells were recorded, and the distribution pattern of the outliers was analyzed. Finally, based on the above data, a preliminary mesh map of stress distribution is generated. A visualization tool is used to map the stress values ​​to a color gradient, setting the colors from blue to red for stress values ​​ranging from 0 to 200 MPa, thus generating a three-dimensional heat map. This visually displays the contrast between stress concentration areas and normal areas, providing data support for subsequent material defect analysis. All these steps are closely linked through algorithms and numerical simulations, ensuring a complete logical chain from sound wave propagation to the generation of the stress distribution map.

[0048] S5: Based on the preliminary grid diagram of stress distribution, obtain the internal stress non-uniformity distribution diagram.

[0049] It should be noted that, based on the preliminary grid map of stress distribution, the local gradient change rate of stress distribution is calculated to quantify the non-uniformity. If the distribution density of stress concentration points is higher than the preset density threshold, a stress distribution visualization map is drawn based on the map generation rules to obtain the internal stress non-uniformity distribution map.

[0050] In this embodiment of the application, obtaining the internal stress non-uniformity distribution map in step S5 includes steps D1-D5: D1: By extracting stress distribution data from the preliminary grid diagram of stress distribution, the characteristics of local stress distribution are analyzed. D2: Calculate and determine the non-uniformity distribution based on the characteristics of local stress distribution; D3: For the results of non-uniform distribution, analyze the distribution density of stress concentration points. If the distribution density is higher than the preset threshold, trigger the subsequent visualization processing to obtain the density determination result. D4: Extract high-density area data from the density determination results and generate preliminary stress visualization graphics; D5: Obtain preliminary stress visualization graphics and generate the final internal stress non-uniformity distribution diagram.

[0051] In this embodiment of the application, the triggering of the subsequent visualization processing flow in step D3 is as follows: A heatmap algorithm is used to color-map the stress values ​​of each mesh element according to the color gradient; An interpolation algorithm is introduced to smooth the color distribution and generate a heat map of the basic stress distribution; By combining the gradient matrix and stress concentration point information, layers are overlaid on the heat map; Regions with local gradients greater than 30 MPa are highlighted with black borders to obtain the final internal stress non-uniformity distribution map.

[0052] It should be noted that the specific implementation methods for steps D1-D3 are as follows: Stress distribution data is extracted from the initial grid map, and each region is analyzed independently using a block-based processing approach to obtain local stress distribution characteristics. Based on these characteristics, the gradient change value for each region is calculated, and gradient analysis is used to quantify the degree of non-uniformity, determining the non-uniformity distribution result. For the non-uniformity distribution result, the distribution density of stress concentration points is analyzed. If the distribution density exceeds a preset threshold, subsequent visualization processing is triggered to obtain density determination results. High-density region data is extracted from the density determination results, and these regions are graphically mapped based on preset map rules to generate a preliminary stress visualization graphic. This preliminary stress visualization graphic is then combined with internal stress data for secondary correction processing to generate the final internal stress non-uniformity distribution diagram. Using this final internal stress non-uniformity distribution diagram, the relationship between local regions and the overall distribution is analyzed. Cluster analysis is used to classify the non-uniform regions, obtaining classification distribution results. Based on the classification distribution results, a targeted analysis file is generated, recording the characteristic data of each type of non-uniform region, completing a comprehensive analysis of the stress distribution.

[0053] Furthermore, assuming preliminary mesh data is obtained from an engineering structural analysis project, the mesh is represented as a 100x100 two-dimensional matrix. Each mesh cell stores the corresponding stress value in megapascals (MPa), ranging from 0 to 500. First, the local gradient rate of change of stress distribution is calculated to quantify non-uniformity. Specifically, each mesh cell is traversed, and the stress difference between it and its eight neighboring cells is calculated. The largest difference is taken as the local gradient. For example, if a cell has a stress value of 300 MPa and the maximum surrounding stress is 350 MPa, the gradient is 50 MPa. The gradient matrix for the entire image is calculated sequentially, and the percentage of cells with gradients greater than 30 MPa is counted. Assuming the result is 25%, this indicates high non-uniformity. Next, the distribution density of stress concentration points is determined. Stress concentration points are defined as cells with stress values ​​greater than 400 MPa, and their number is counted as 200. The total mesh area is 10,000 cells, and the density is 0.02. If the preset density threshold is 0.015, values ​​exceeding the threshold trigger the visualization process. Based on map generation rules, a heatmap algorithm is used to map stress values ​​to color gradients: values ​​above 400 MPa are set to red, 200 to 400 MPa to yellow, and below 200 MPa to green. An interpolation algorithm smooths the color transitions, generating a stress distribution visualization. Finally, combining the gradient matrix and the distribution of concentration points, regions with gradients greater than 30 MPa are overlaid and marked with black borders to obtain an internal stress non-uniformity distribution map. Analysis results show that concentration points are mostly distributed in the center of the grid, coinciding with high gradient value areas, suggesting potential structural defects. Further verification using material parameters is needed, forming a complete logical chain from data processing to result analysis.

[0054] S6: By aligning the preliminary distribution matrix of surface material composition with the internal stress non-uniformity distribution map in spatial position, comprehensive quality risk distribution data is obtained.

[0055] It should be noted that by aligning the preliminary distribution matrix of the surface material composition with the internal stress non-uniformity distribution map in terms of spatial location, and analyzing the correlation between the component correlation coefficient and the local gradient change rate of the stress distribution for the spatial location marked in the high-risk area, comprehensive quality risk distribution data is obtained.

[0056] In this embodiment of the application, obtaining the comprehensive quality risk distribution data in step S6 includes steps E1-E3: E1: By reading the preliminary distribution matrix of surface materials and their components, spatial location information in the initial matrix is ​​obtained, and the component data corresponding to each location point is determined.

[0057] E2: Based on the obtained spatial location information, the internal stress is matched with the non-uniformity distribution pattern to obtain the stress distribution data corresponding to the initial matrix.

[0058] E3: For high-risk areas and marked locations, extract the corresponding component data and stress distribution data, calculate the correlation value between the two in spatial location, and determine the comprehensive quality risk distribution data based on the correlation value.

[0059] In this embodiment of the application, the determination of the comprehensive quality risk distribution data based on the correlation value in step E3 is as follows: If the correlation value exceeds the preset correlation value threshold range, a local analysis of the gradient change in the high-risk area is performed to determine whether there are abnormal fluctuation areas. By analyzing the correlation between abnormal fluctuation areas and quality risks, the risk level is classified using the support vector machine algorithm, and the risk distribution results for each area are obtained. Based on the classification results, a corresponding quality risk distribution view is generated for the gradient change data between high-risk areas and marked locations, and the final risk assessment data is determined. If outliers are found in the risk assessment data, a secondary verification is performed by combining the spatial location and the non-uniform distribution graph to obtain corrected quality risk distribution data.

[0060] Furthermore, the specific implementation methods for steps E1-E3 are as follows: By reading the initial matrix of surface material and composition distribution, spatial location information is obtained from the initial matrix, and the composition data corresponding to each location point is determined. Based on the obtained spatial location information, the internal stress and non-uniformity distribution pattern are matched to obtain stress distribution data corresponding to the initial matrix. For high-risk areas and marked locations, the corresponding composition data and stress distribution data are extracted, and the correlation value between the two in spatial location is calculated. If the correlation value exceeds the preset correlation value threshold range, the gradient change in the high-risk area is locally analyzed to determine whether there are abnormal fluctuation areas. Through the correlation analysis between abnormal fluctuation areas and quality risk, the risk level is classified using the support vector machine algorithm to obtain the risk distribution results for each area. Based on the classification results, the gradient change data of high-risk areas and marked locations are used to generate the corresponding quality risk distribution view, and the final risk assessment data is determined. If there are outliers in the risk assessment data, a secondary verification is performed by combining the spatial location and non-uniformity distribution pattern to obtain the corrected quality risk distribution information.

[0061] Furthermore, when aligning the spatial positions of the preliminary surface material composition distribution matrix and the internal stress non-uniformity distribution map, image processing algorithms can be used to uniformly calibrate their coordinate systems. Assuming the resolution of the surface material distribution matrix is ​​100×100 pixels, with each pixel representing an actual distance of 0.1 mm, while the resolution of the stress distribution map is 200×200 pixels, with each pixel representing an actual distance of 0.05 mm, an interpolation algorithm is first used to unify the two to a resolution of 0.05 mm / pixel. Then, a feature point matching algorithm (such as SIFT) is used to extract key points from the two images, calculate the spatial transformation matrix, and achieve pixel-level alignment with an error controlled within ±0.01 mm. Next, regarding the spatial location of the high-risk areas, assuming they lie within the matrix coordinates (50,50) to (70,70), the correlation coefficients of components within this area (e.g., the correlation coefficient between iron content and carbon content is 0.85) and the local gradient change rate of stress distribution (calculated using the Sobel operator, yielding a change rate of 0.3 MPa / mm) were extracted. The Pearson correlation analysis method was used to calculate the correlation between these two factors, yielding a correlation coefficient of 0.72, indicating a strong positive correlation between component distribution and stress change. Subsequently, the correlation analysis results were combined with the stress peak value of the high-risk areas (e.g., a maximum stress of 500 MPa), and a weighted calculation model (with weights of 0.6 and 0.4) was used to generate comprehensive quality risk distribution data. The risk value ranges from 0 to 1, with the risk value at the center point (60,60) of the high-risk area being 0.88, indicating a high probability of quality problems in this area. To form a logical chain, risk distribution data can be combined with a material fatigue life prediction model. Assuming that the fatigue life in areas with a risk value greater than 0.8 is shortened to 60% of its original life (i.e., from 1 million cycles to 600,000 cycles), a risk warning report is automatically generated by an algorithm for subsequent process optimization. All of the above processes are automated through computer programs, ensuring the objectivity and efficiency of data analysis.

[0062] S7: Based on the comprehensive quality risk distribution data, a comprehensive judgment is made to obtain the optimized processing control sequence.

[0063] It should be noted that, based on the comprehensive quality risk distribution data, if the proportion of abnormal areas in the risk distribution data exceeds the preset range, the processing parameter adjustment mechanism is triggered to fine-tune the local temperature field distribution during the processing, and the spectral signal and ultrasonic echo signal are re-acquired to determine the surface composition stability and internal stress uniformity distribution after adjustment, thereby obtaining the optimized processing control sequence.

[0064] Furthermore, comprehensive quality risk distribution data is acquired. By performing stratified analysis on this distribution data, the proportion of abnormal areas is determined to be within a preset range, thus obtaining the distribution status of abnormal areas. If the distribution status of abnormal areas exceeds the preset range, a parameter adjustment process is triggered to dynamically correct the local temperature distribution during processing, determining the corrected temperature distribution data. Based on the corrected temperature distribution data, spectral signals and ultrasonic echo signals during processing are re-acquired. Signal processing techniques are used to denoise and extract features from the acquired signals, resulting in a clear set of signal features. For this set of signal features, a support vector machine algorithm is used to classify and evaluate the stability of the surface composition, determining whether the surface composition meets the preset stability standard, thus obtaining the evaluation result of the surface composition. Based on the evaluation result of the surface composition, combined with the feature data of the ultrasonic echo signal, spatial distribution analysis of internal stress uniformity is performed to determine the consistency status of internal stress distribution. Based on the consistency status of internal stress distribution, an optimized processing control sequence is generated, and key parameters during processing are updated in real time to obtain the final control sequence data. Using the final control sequence data, the processing process is continuously monitored and parameters are fine-tuned to determine whether quality risks are effectively controlled, thus obtaining the stable operating status of the processing process.

[0065] Furthermore, in processing the comprehensive quality risk distribution data, the system first analyzes the risk distribution data during the processing. Assuming a total of 10,000 data points are collected, with 1,500 data points in abnormal areas (15%), and the preset range being 10%, the system automatically determines that the threshold is exceeded, triggering a processing parameter adjustment mechanism. Next, the system fine-tunes the local temperature field distribution during processing, employing a temperature field optimization algorithm based on finite element analysis. The initial temperature field non-uniformity is ±5°C. By adjusting the heating power and cooling rate, the non-uniformity is reduced to ±2°C. Specifically, the heating power is increased from 500W to 550W, and the cooling rate is adjusted from 2°C / s to 1.5°C / s. Simulation results show that the variance of the temperature field distribution decreases from 2.5 to 1.2. Subsequently, the system reacquired spectral and ultrasonic echo signals, and used spectral analysis algorithms to evaluate the surface composition stability. Assuming the spectral peak shift decreased from 0.05 nm to 0.02 nm, it indicated a 60% improvement in composition stability. Simultaneously, through time-domain analysis of the ultrasonic echo signals, the internal stress uniformity distribution was calculated. The initial stress distribution standard deviation was 10 MPa, which was adjusted to 6 MPa, representing a 40% improvement in uniformity. Finally, based on the above data, the system generated an optimized processing control sequence and used a genetic algorithm to iteratively optimize the processing parameters. The number of iterations was set to 100, and the objective function was to minimize the stress distribution standard deviation and temperature field non-uniformity. The final output control sequence stabilized the heating power at 545 W and the cooling rate at 1.48°C / s. The optimization results were stored in the database, forming a closed-loop control logic to ensure the stability of subsequent processing. To further strengthen the logic chain, the system also linked to the processing equipment's operating status monitoring module, collecting real-time equipment vibration frequency data. Assuming the vibration frequency decreased from 5 Hz to 3 Hz, it indicated improved equipment stability, creating a synergistic effect with the processing parameter adjustments to ensure overall processing quality.

[0066] It should be noted that, in this document, the terms "comprising," "including," and any other variations are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Specific examples have been used in this document to illustrate the principles and implementation methods of the present invention. These examples are merely for the purpose of helping to understand the method and core ideas of the present invention. The above descriptions are only preferred embodiments of the present invention. It should be pointed out that, due to the limitations of written expression and the objective existence of infinite specific structures, those skilled in the art can make several improvements, modifications, or variations without departing from the principles of the present invention, and can also combine the above technical features in an appropriate manner. These improvements, modifications, variations, or combinations, or the direct application of the concept and technical solution of the present invention to other situations without modification, should all be considered within the scope of protection of the present invention.

[0067] Example 3, as Figure 2 As shown, this embodiment provides an intelligent quality measurement system for a product manufacturing process, including: The spectral scanning module continuously scans the product surface using a spectral analyzer to obtain a preliminary distribution matrix of the surface material composition. The composition labeling module processes the preliminary distribution matrix of the surface material composition to obtain labeled risk distribution data; The waveform extraction module extracts the spatial location of high-risk areas from the marked risk distribution data to obtain the original waveform set of internal stress distribution. The stress mesh module generates a preliminary mesh diagram of the stress distribution based on the original waveform set of the internal stress distribution. The non-uniformity analysis module obtains an internal stress non-uniformity distribution map based on the preliminary mesh map of stress distribution; The spatial alignment module aligns the preliminary distribution matrix of surface material composition with the internal stress non-uniformity distribution map in spatial position to obtain comprehensive quality risk distribution data. The processing optimization module makes a comprehensive judgment based on the overall quality risk distribution data to obtain the optimized processing control sequence.

[0068] It should be noted that, in this document, the terms "comprising," "including," and any other variations are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Specific examples have been used in this document to illustrate the principles and implementation methods of the present invention. These examples are merely for the purpose of helping to understand the method and core ideas of the present invention. The above descriptions are only preferred embodiments of the present invention. It should be pointed out that, due to the limitations of written expression and the objective existence of infinite specific structures, those skilled in the art can make several improvements, modifications, or variations without departing from the principles of the present invention, and can also combine the above technical features in an appropriate manner. These improvements, modifications, variations, or combinations, or the direct application of the concept and technical solution of the present invention to other situations without modification, should all be considered within the scope of protection of the present invention.

Claims

1. An intelligent quality measurement method for a product manufacturing process, characterized in that, include: By continuously scanning the product surface with a spectrometer, a preliminary distribution matrix of the surface material composition is obtained; By processing the preliminary distribution matrix of the surface material composition, the labeled risk distribution data is obtained; The spatial locations of high-risk areas are extracted from the labeled risk distribution data to obtain the original waveform set of internal stress distribution; Based on the original waveform set of the internal stress distribution, a preliminary mesh diagram of the stress distribution is obtained; Based on the preliminary grid diagram of stress distribution, the internal stress non-uniformity distribution diagram is obtained; By aligning the preliminary distribution matrix of surface material composition with the internal stress non-uniformity distribution map in terms of spatial location, comprehensive quality risk distribution data is obtained. Based on the comprehensive quality risk distribution data, a comprehensive judgment is made to obtain the optimized processing control sequence.

2. The intelligent quality measurement method for a product manufacturing process as described in claim 1, characterized in that: The obtained labeled risk distribution data includes: An initial spatiotemporal feature mapping matrix is ​​constructed to obtain preliminary spectral signal distribution data; Based on the preliminary spectral signal distribution data, the fluctuation amplitude data of the signal at different time points and spatial locations are determined; If the fluctuation amplitude data exceeds the preset fluctuation amplitude threshold range, then obtain the distribution data of the marked abnormal areas; Based on the marked abnormal region distribution data, determine whether the abnormal region belongs to the high-risk category to obtain the classified risk region data; The specific location information of high-risk areas is extracted from the classified risk area data. By comparing it with the spatiotemporal feature mapping matrix, the corresponding range of high-risk areas in the dynamic matrix is ​​determined, and the spatiotemporal distribution data of high-risk areas is obtained. In-depth analysis of the peak characteristics of the spatiotemporal distribution data spectral signals of high-risk areas is conducted to obtain peak characteristic distribution data; By using peak feature distribution data, high-risk areas are finally labeled to generate complete risk distribution data.

3. The intelligent quality measurement method for a product manufacturing process as described in claim 1, characterized in that: The original waveform set for obtaining the internal stress distribution includes: By extracting the spatial location of high-risk areas from risk distribution data, a preliminary echo data set is obtained; Based on the acquired preliminary echo data set, data cleaning is performed to obtain a cleaned waveform signal group; For the cleaned waveform signal group, determine the corrected waveform dataset; Preliminary mapping data of stress distribution are obtained from the corrected waveform dataset; By using preliminary mapping data of stress distribution, potential areas of high stress concentration can be identified; If the classification results show that the high stress concentration area exceeds the preset stress threshold, then the original waveform data of the current area will be analyzed a second time. Based on the stress distribution characteristics obtained from the secondary analysis, a detailed stress distribution map of the corresponding high-risk area is generated, the final detection result is determined, and the original waveform set of the internal stress distribution is generated.

4. The intelligent quality measurement method for a product manufacturing process as described in claim 1, characterized in that: The preliminary mesh diagram of the obtained stress distribution includes: Based on the waveform data related to internal stress, simulation results of the propagation path are obtained; Based on the simulation results of the propagation path, spatial units are divided to determine the potential regions of stress anomalies; For potential areas of stress anomalies, obtain the specific coordinate information of the concentration points; By using the specific coordinate information of the concentration points and the results of spatial unit division, a preliminary grid diagram of stress distribution is constructed, and the distribution pattern of stress anomalies in the grid diagram is determined. Based on the distribution pattern of stress anomalies in the mesh diagram, a preliminary mesh diagram of stress distribution is generated.

5. The intelligent quality measurement method for a product manufacturing process as described in claim 4, characterized in that: The determination of the distribution pattern of stress anomalies in the mesh diagram includes: If the stress anomaly distribution in the mesh diagram shows a concentrated trend, then the spatial units around the concentrated points are refined into a finer mesh to obtain the stress anomaly boundary and thus the refined mesh distribution data. Based on the refined grid distribution data, the support vector machine algorithm is used to classify the abnormal areas, distinguish different levels of stress concentration, and determine the stress distribution map. Analyze the correlation between stress anomalies and concentration points within spatial cells in the stress distribution map to generate detailed mapping results of stress distribution.

6. The intelligent quality measurement method for a product manufacturing process as described in claim 1, characterized in that: The obtained internal stress non-uniformity distribution map includes: By extracting stress distribution data from the preliminary grid diagram of stress distribution, the characteristics of local stress distribution are analyzed. Based on the characteristics of local stress distribution, the non-uniform distribution results are calculated and determined; For the results of non-uniform distribution, the distribution density of stress concentration points is analyzed. If the distribution density is higher than a preset threshold, the subsequent visualization processing is triggered to obtain the density determination result. Data from high-density areas are extracted from the density determination results to generate preliminary stress visualization graphics; Obtain preliminary stress visualization graphics and generate the final internal stress non-uniformity distribution diagram.

7. The intelligent quality measurement method for a product manufacturing process as described in claim 6, characterized in that: The triggering of subsequent visualization processing includes: A heatmap algorithm is used to color-map the stress values ​​of each mesh element according to the color gradient; An interpolation algorithm is introduced to smooth the color distribution and generate a heat map of the basic stress distribution; By combining the gradient matrix and stress concentration point information, layers are overlaid on the heat map; Regions with local gradients greater than 30 MPa are highlighted with black borders to obtain the final internal stress non-uniformity distribution map.

8. The intelligent quality measurement method for a product manufacturing process as described in claim 1, characterized in that: The obtained comprehensive quality risk distribution data includes: By reading the preliminary distribution matrix of surface materials and their components, spatial location information in the initial matrix is ​​obtained, and the component data corresponding to each location point is determined. Based on the obtained spatial location information, the internal stress is matched with the non-uniform distribution pattern to obtain stress distribution data corresponding to the initial matrix; For high-risk areas, the corresponding component data and stress distribution data are extracted, and the correlation value between the two in spatial location is calculated. Based on the correlation value, the comprehensive quality risk distribution data is determined.

9. The intelligent quality measurement method for a product manufacturing process as described in claim 8, characterized in that: The process of determining the comprehensive quality risk distribution data based on correlation values ​​includes: If the correlation value exceeds the preset threshold range, a local analysis of the gradient changes in the high-risk area is performed to determine whether there are abnormal fluctuation areas. By analyzing the correlation between abnormal fluctuation areas and quality risks, the risk level is classified using the support vector machine algorithm, and the risk distribution results for each area are obtained. Based on the classification results, a corresponding quality risk distribution view is generated for the gradient change data of the marked locations in high-risk areas, and the final risk assessment data is determined. If outliers are found in the risk assessment data, a secondary verification is performed by combining the spatial location and the non-uniform distribution graph to obtain corrected quality risk distribution data.

10. An intelligent quality measurement system for a product manufacturing process, employing the method described in any one of claims 1-9, characterized in that, include: The spectral scanning module continuously scans the product surface using a spectral analyzer to obtain a preliminary distribution matrix of the surface material composition. The composition labeling module processes the preliminary distribution matrix of the surface material composition to obtain labeled risk distribution data; The waveform extraction module extracts the spatial location of high-risk areas from the marked risk distribution data to obtain the original waveform set of internal stress distribution. The stress mesh module generates a preliminary mesh diagram of the stress distribution based on the original waveform set of the internal stress distribution. The non-uniformity analysis module obtains an internal stress non-uniformity distribution map based on the preliminary mesh map of stress distribution; The spatial alignment module aligns the preliminary distribution matrix of surface material composition with the internal stress non-uniformity distribution map in spatial position to obtain comprehensive quality risk distribution data. The processing optimization module makes a comprehensive judgment based on the overall quality risk distribution data to obtain the optimized processing control sequence.

Images