Stainless steel rotary body welding seam detecting and grinding method

Abstract

The invention provides a stainless steel rotary body welding seam detecting and grinding method which comprises the steps that if the severity degree of a defect reaches a preset threshold value standard, a defect triggering instruction is generated, a production halt signal is sent to related equipment through a production halt control mechanism, and production halt execution state feedback is obtained; through a polishing path optimization scheme, accurate positioning and polishing operation are conducted on the defect area, surface image data in the polishing process are collected in real time, and a dynamic monitoring result of the polishing effect is obtained; according to the dynamic monitoring result of the polishing effect, secondary evaluation is conducted on the surface quality of the weld joint through a surface quality monitoring mechanism, and whether polishing meets the preset quality standard or not is judged; and if the polishing effect does not meet the preset quality standard, the polishing path optimization scheme is adjusted again, new polishing parameters and trajectory data are obtained, and polishing operation is executed circularly.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to a method for detecting and polishing welds of a stainless steel rotating body. Background Art

[0002] Utilizing industrial cameras and image processing algorithms, online visual inspection of weld surface quality is performed. When surface defects are found, a production stop instruction is triggered, enabling rapid identification and location of defects. Through visual production stop technology, comprehensive monitoring of weld surface quality can be achieved, improving defect detection rate and recognition accuracy. At the same time, the grinding plan and path can be optimized, improving grinding efficiency and quality, and achieving refined processing of the weld surface. Summary of the Invention

[0003] The present invention provides a stainless steel rotating body weld detection and grinding method, which mainly includes: The industrial camera equipment is used to collect real-time images of the weld surface of the stainless steel rotating body, and the collected raw image data is preliminarily stored and format converted to obtain weld surface image data for subsequent processing; Image processing algorithms are used to perform denoising and edge enhancement on weld surface image data, and feature extraction is performed on the processed images to determine the distribution of potential defect areas on the weld surface. Based on the distribution of potential defects, a pre-established defect classification model is used to analyze the defect type and severity to determine whether the defects have reached the preset threshold standard that triggers a production stoppage. If the severity of the defect reaches the preset threshold, a defect trigger instruction is generated, and a shutdown signal is sent to the relevant equipment through the shutdown control mechanism to obtain shutdown execution status feedback; Based on the feedback from the production stoppage execution status, an initial plan for grinding path optimization is generated based on the specific location and characteristics of the defect area, and the movement trajectory and parameter settings of the grinding tool are determined; Through the polishing path optimization solution, the defect area is accurately located and polished, and the surface image data of the polishing process is collected in real time to obtain dynamic monitoring results of the polishing effect; Based on the dynamic monitoring results of the grinding effect, the surface quality monitoring mechanism is used to conduct a secondary assessment of the weld surface quality to determine whether the grinding meets the preset quality standards; If the polishing effect does not meet the preset quality standards, the polishing path optimization plan is readjusted, new polishing parameters and trajectory data are obtained, and the polishing operation is executed cyclically; After the polishing effect meets the preset quality standards, the final surface quality monitoring data and defect identification accuracy information are recorded, and a complete inspection and polishing report is generated and stored in the system database.

[0004] The technical scheme provided by the embodiment of the present application can include the following beneficial effects: The welding seam surface image is collected in real time by the industrial camera, the image processing algorithm is used for denoising and edge enhancement, the feature is extracted to determine the potential defect area, and the defect type and severity are analyzed according to the preset model. When the defect reaches the preset threshold, the present application automatically triggers the production stop and generates a polishing path optimization scheme, and the defect area is accurately positioned and polished. The surface quality is monitored in real time during the polishing process, and if it does not meet the standard, the polishing parameters are dynamically adjusted until the preset standard is met. The present application realizes automatic detection, intelligent production stop and accurate polishing of the welding seam surface defects, improves the production efficiency and product quality, reduces the manual intervention, and provides an efficient and reliable quality guarantee scheme for the stainless steel rotary body manufacturing. BRIEF DESCRIPTION OF DRAWINGS

[0005] Figure 1 A flowchart of a stainless steel rotary body welding seam detection and polishing method.

[0006] Figure 2 A schematic diagram of a stainless steel rotary body welding seam detection and polishing method.

[0007] Figure 3 Another schematic diagram of a stainless steel rotary body welding seam detection and polishing method. DETAILED DESCRIPTION

[0008] In order to further understand the content of the present application, the present application will be described in detail in conjunction with the drawings and embodiments. The specific embodiments described herein are only used to explain the related application, and are not limited to the application. In addition, it should be noted that, in order to facilitate description, only the parts related to the application are shown in the drawings.

[0009] As Figure 1-3 , the stainless steel rotary body welding seam detection and polishing method can specifically include: Step S101, the welding seam surface of the stainless steel rotary body is collected in real time by the industrial camera device, the collected original image data is stored and converted in format, and the welding seam surface image data for subsequent processing is obtained.

[0010] An industrial camera acquires real-time images of the weld surface of a stainless steel rotating body at a preset acquisition frequency to obtain raw image data. If the format of the raw image data does not meet preset processing requirements, a data conversion tool is used to convert the raw image data into a target format to obtain first image data. Based on storage capacity limitations, the first image data is initially stored using a compression algorithm to obtain compressed image data. The compressed image data is then acquired and image quality is adjusted using an image enhancement algorithm to obtain weld surface image data.

[0011] Specifically, real-time image capture of the weld surface of a stainless steel rotating body was performed using an industrial camera. An industrial CCD camera with a resolution of 1920×1080 pixels and a frame rate of 30 fps was used to capture dynamic details of the weld surface. The lens had a focal length of 25 mm and a working distance of 200 mm to ensure clear imaging. The captured raw image data was in RAW format, with a single frame size of approximately 2.07 MB (1920×1080×12 bits). The data was transmitted to a storage server via Gigabit Ethernet. Using the real-time Zstandard compression algorithm, the data was compressed to approximately 1.2 MB per frame and stored on an NVMe SSD with a bandwidth of 3.5 GB / s, meeting real-time storage requirements. Approximately 36 MB of data was stored per second, ensuring continuous acquisition without frame loss. The captured RAW images required format conversion for subsequent processing, which was implemented using the OpenCV library. First, white balance correction was performed on the RAW images. Based on the grayscale world algorithm, assuming an average grayscale value of 128, the RGB channel gains were adjusted to achieve color balance. The RAW data is then demosaiced using a Bayer filter, converting it to RGB format. The resolution remains at 1920×1080, with a pixel depth of 8 bits per channel, resulting in an RGB image of approximately 6.22MB per frame. To optimize storage and processing efficiency, JPEG compression (90% compression quality) is applied to compress the RGB image to approximately 0.8MB per frame. The image is then stored in the HDFS distributed file system to ensure high data availability. The converted image data undergoes edge detection preprocessing using the Canny algorithm with a low threshold of 50 and a high threshold of 150 to extract weld edge features and generate a binary image with unchanged resolution for subsequent defect detection. The entire process is controlled by an automated script, implemented in Python and OpenCV. Image acquisition, storage, format conversion, and preprocessing are seamlessly integrated, with processing latency kept to under 50ms, ensuring real-time performance and a closed-loop data flow that meets the business requirements of weld surface quality analysis.

[0012] Step S102: Using an image processing algorithm to perform denoising and edge enhancement on the weld surface image data, performing feature extraction on the processed image, and determining the distribution of potential defect areas on the weld surface.

[0013] Image processing technology is used to perform preliminary processing on the weld surface image data, and a denoising operation is performed to obtain the first image data. Edge enhancement technology is implemented on the first image data to obtain the second image data. Based on the second image data, key feature information is extracted to determine the distribution of significant features on the weld surface. By analyzing the distribution of significant features, if the feature distribution deviates from the preset threshold range, it is determined that there is a potential defect area and preliminary defect location information is obtained. Based on the preliminary defect location information, combined with the spatial distribution characteristics of the weld surface, the convolutional neural network in the deep learning model is used for refined classification to obtain a specific classification of defect types. Based on the specific classification of defect types, the defect area is spatially mapped to determine the precise distribution range of the defect area. By integrating data on the precise distribution range of the defect area, comprehensive distribution information of weld surface defects is generated.

[0014] Specifically, image denoising techniques are first employed to improve image quality. A Gaussian filter algorithm is used. The filter kernel size is set to 5x5 and the standard deviation is set to 1.5. The original image is smoothed to reduce noise. Assuming the original image grayscale noise variance is 20, after filtering, the noise variance is reduced to below 5, preserving image detail. Edge enhancement is then performed. The Sobel operator is used to calculate horizontal and vertical gradients, with a threshold of 100. Pixels with gradient magnitudes greater than this threshold are marked as edges, enhancing the saliency of features such as cracks or bumps on the weld surface. After processing, the grayscale contrast of edge pixels increases by approximately 30%. Feature extraction is then performed on the processed image. The Canny edge detection algorithm is used with a low threshold of 50 and a high threshold of 150 to extract continuous edge features on the weld surface. Morphological operations such as dilation and erosion (with a 3x3 structuring element) are then applied to fill in small fracture edges. Assuming that five regions with edge lengths greater than 10 pixels are detected after processing, these regions are marked as potential defect candidates. Finally, by analyzing the geometric features of the candidate areas, areas with an area greater than 50 square pixels and an aspect ratio greater than 2 are determined to be potential defect areas. Assume that three defect areas are finally identified, located near the image coordinates (100, 200), (300, 400), and (500, 150), respectively. By calculating the difference between the grayscale mean of the defect area and the grayscale mean of the surrounding background (assuming the difference is greater than 20), the severity of the defect is further confirmed, forming a complete logical chain from denoising to defect location. If the number of candidate areas in feature extraction is insufficient, the weld surface texture analysis can be introduced, and the contrast features can be calculated in combination with the grayscale co-occurrence matrix (assuming the contrast threshold is 0.5) to assist in screening potential defects and ensure comprehensive analysis and reliable results.

[0015] Step S103 , based on the distribution of potential defect areas, a pre-established defect classification model is used to analyze the defect type and severity to determine whether the defect reaches a preset threshold standard that triggers a production stop.

[0016] Obtain the distribution data of potential defect areas, standardize and clean the data through the preprocessing module to obtain a structured defect distribution data set. Use the pre-established random forest model to classify the structured defect distribution data set, analyze the defect types, and obtain defect type labels. Use the support vector machine model to evaluate the severity of the defect type labels and generate a defect severity score. If the defect severity score exceeds the preset threshold, the logistic regression model is used to predict the probability of production suspension and obtain the possibility of production suspension triggering. Based on the production suspension triggering possibility, combined with the distribution feature analysis module, high-risk defect areas are extracted and a defect area priority list is generated. Based on the defect area priority list, the automated decision-making module is called to determine whether the production suspension criteria are met and generate a production suspension decision instruction. Based on the production suspension decision instruction, the production monitoring database is updated, the defect processing status is recorded, and the basis for production adjustment is obtained.

[0017] Specifically, to analyze and process the distribution of potential defect areas, image processing technology is first used to detect defect areas in part surface images collected on the production line. Assuming the image resolution is 1920x1080 pixels, a deep learning-based convolutional neural network model (such as ResNet-50) is used to segment the image and identify the coordinates and area of ​​the defect area. Assuming a defect area of ​​12.5 square millimeters is detected, located slightly to the left of the image center (coordinates x=800, y=540). Next, the identified defect area data is input into a pre-established defect classification model. This model uses a support vector machine algorithm (SVM) trained with historical data. Features include defect area, shape factor, and grayscale value. Assuming the shape factor of the defect is 0.75 and the grayscale value is 120, the model output is a "crack" type with a severity score of 7.2 (out of 10), indicating that the defect is relatively serious. The system then makes a judgment based on preset threshold criteria. Assuming the production suspension threshold is a severity greater than 7.0 and a defect area greater than 10.0 square millimeters, the current defect meets the conditions (7.2 > 7.0 and 12.5 > 10.0), triggering a production suspension warning. The system automatically associates the analysis results with the production batch data. Assuming the batch number is A20231001, the defect data is recorded in the database, and a production suspension instruction is generated and sent to the production control system. The quality management module is also notified for subsequent traceability analysis, ensuring that the defect is promptly addressed, forming a complete closed-loop logic from detection to decision-making.

[0018] In step S104, if the severity of the defect reaches a preset threshold, a defect trigger instruction is generated, and a stop signal is sent to relevant equipment through the stop control mechanism to obtain stop execution status feedback.

[0019] The defect detection system monitors defect severity in real time, comparing it to preset thresholds to determine whether trigger conditions have been met. If the defect severity exceeds the preset threshold, the command generation logic constructs a defect trigger command, determining the command's priority and target device. Based on the output of the command generation logic, the shutdown control mechanism sends a shutdown signal to the relevant devices and obtains the transmission status of the signal transmission path. After obtaining the transmission status from the signal transmission path, the successful delivery of the shutdown signal is determined based on the reception of the relevant device signals, and the device response information is obtained. If the device response indicates that the shutdown signal was not successfully delivered, the feedback processing mechanism regenerates the shutdown signal and determines an adjustment plan for the signal transmission path. Based on the adjusted signal transmission path, the shutdown signal is resent, and the shutdown execution status is obtained to determine whether the device's operating status meets expectations. If the device's operating status meets expectations, the shutdown execution status is recorded through status feedback information, generating complete execution log data.

[0020] Specifically, in the defect detection and control process, assume the system uses image recognition technology to detect surface defects on products on the production line. The threshold for defect severity is set at a defect area exceeding 5%. In implementation, the system uses a high-resolution camera to capture 10 product images per second and uses a deep learning algorithm to calculate the ratio of the defect area to the total area. For example, if a product has a defect area of ​​2.5 square centimeters and a total area of ​​50 square centimeters, the ratio is 5%, which exactly meets the threshold. The system automatically determines it as a severe defect and generates a defect trigger instruction. The system then sends a stop signal to the relevant equipment via the Industrial Internet of Things protocol, such as the assembly line master control programmable logic controller (PLC). Signal transmission delay is controlled within 0.1 seconds to ensure real-time performance. The system also records the signal transmission timestamp and device ID to create a traceable log. Subsequently, after the stop control mechanism is activated, the equipment returns execution status feedback. For example, if the PLC completes the stop operation within 0.2 seconds of receiving the signal, it returns a status code of "1" indicating successful execution or "0" indicating failure via the data interface. After receiving feedback, if the status code is "0," the system automatically triggers the backup shutdown mechanism, sending a secondary shutdown signal to the backup control module to ensure production safety. The entire process utilizes algorithms to analyze defect data, automatically transmit commands, and provide status feedback, forming a closed-loop control logic to ensure efficient and reliable defect handling. Furthermore, to enhance system stability, the system automatically analyzes the number of shutdowns every hour. If it exceeds 10, an optimization algorithm is triggered, adjusting the detection threshold to 6% to reduce false positives and ensure production continuity.

[0021] Step S105 , based on the feedback of the production stoppage execution status, an initial plan for optimizing the grinding path is generated according to the specific location and characteristics of the defective area, and the movement trajectory and parameter settings of the grinding tool are determined.

[0022] Data related to defective areas is obtained from execution feedback from production stoppages. Preliminary analysis is performed on the feedback information, extracting descriptive information about specific locations and regional characteristics, resulting in a preliminary defect distribution mapping. Based on this preliminary defect distribution mapping, the specific locations and regional characteristics are classified, and defect severity is graded using a preset threshold to determine the priority for each region. Based on the priority order and combined with the data from the specific locations, an initial polishing path plan is generated. The A-star algorithm, part of the path planning algorithm, is used to perform a preliminary calculation of the movement trajectory, resulting in a preliminary path layout. Based on this preliminary path layout, the polishing tool parameters are adjusted based on the regional characteristics and defect severity. If the parameter configuration does not match the preset threshold range, the movement trajectory is locally corrected to determine an optimized path plan. This optimized path plan is dynamically adjusted based on the tool control real-time data. Any abnormal fluctuations detected in the real-time data trigger a recalibration of the parameter configuration, resulting in the final trajectory data. The final trajectory data is then used to verify the integrity of the path plan by comparing it with historical records from the execution feedback. If any deviations are found, the path plan is fine-tuned to determine the final execution plan. Through the final execution plan, control instructions for the grinding tool are generated, the movement trajectory and parameter configuration are monitored in real time, and the status data during the execution process is obtained to determine whether it meets the preset standards.

[0023] Specifically, based on the feedback of the production stoppage execution status, the specific location and feature data of the surface defects of the workpiece are first obtained through the defect detection system. For example, a high-precision laser scanner is used to generate point cloud data with an accuracy of 0.01mm. The coordinates of the defect area are identified as (x1, y1, z1) = (10.5, 20.3, 0.2), and the defect type is a scratch with a depth of 0.3mm and an area of ​​about 2mm². Next, the defect characteristics are analyzed, and the K-means clustering algorithm is used to divide the defect area into 3 categories. The polishing priority is determined according to the depth and area, and the scratch depth > 0.2mm has the highest priority. When generating the polishing path, the movement trajectory is optimized based on the A* algorithm, the grid resolution is set to 0.5mm, the starting point is the initial position of the tool (0, 0, 0), the end point is the defect center (10.5, 20.3, 0.2), avoiding the processed area, and the shortest path length is calculated to be 25.7mm. During path planning, the tool moves perpendicular to the surface along the Z axis at a speed of 50 mm / s, with a track spacing of 0.1 mm to ensure coverage of the defect area. Grinding parameters are set based on defect depth and material hardness (for example, HB60 for aluminum alloy). A grinding wheel grit size of P120 and a rotation speed of 12,000 rpm are selected, with a grinding force controlled at 5 N. A PID algorithm is used to adjust force deviation, with a proportional coefficient Kp = 0.5 and an integral coefficient Ki = 0.1. After path execution, real-time data from the grinding force sensor is collected to detect force fluctuations within a ±0.5 N range. If these limits are exceeded, the speed is adjusted to 40 mm / s. Finally, a second laser scan verifies the grinding results, confirming that the defect depth has been reduced to <0.05 mm and the surface roughness Ra is <0.8 μm. This algorithm- and data-driven solution ensures efficient paths and precise parameters, meeting the requirements of automated production.

[0024] In step S106 , the defective area is accurately located and polished through the polishing path optimization solution, and surface image data during the polishing process is collected in real time to obtain dynamic monitoring results of the polishing effect.

[0025] The surface image data of the polishing process is acquired in real time via a camera, and image segmentation technology is used to extract defect features and determine the coordinate position of the defect area. Based on the coordinate position of the defect area, the A-star algorithm is used to generate the shortest polishing path and determine the movement trajectory of the polishing equipment. The surface image data is compared with a preset standard image. If the pixel difference between the surface image data and the preset standard image is less than a preset threshold, the polishing effect is judged to be qualified and deviation data is obtained. The force parameters and path trajectory of the polishing equipment are adjusted based on the deviation data. If the deviation data exceeds the preset range, the path is recalculated and the force parameters are updated, and the adjusted effect data is stored.

[0026] In step S107, according to the dynamic monitoring result of the polishing effect, the surface quality monitoring mechanism is used to perform secondary evaluation on the weld surface quality, and it is judged whether the polishing meets the preset quality standard.

[0027] The polishing effect data of the weld surface is obtained by real-time monitoring, the surface quality related information is collected by using the sensor technology, and the preliminary monitoring data set is obtained. According to the preliminary monitoring data set, the feature extraction is performed on the weld surface, the pre-established surface quality feature library is used to determine the initial evaluation result of the surface quality. If the initial evaluation result does not reach the preset standard, the polishing effect of the weld surface is analyzed in depth through the monitoring mechanism, the specific distribution information of the abnormal area is obtained, and it is judged whether there is a local defect. According to the distribution information of the abnormal area, the image processing technology is used to scan the details of the weld surface, and the high-resolution surface quality image data is obtained. Through the high-resolution surface quality image data, the support vector machine algorithm is used to classify the defect types of the weld surface, and the specific category and severity of the defect are determined. If the defect category and severity exceed the preset standard, the data of the defect area is marked, the marked defect data set is obtained, and it is judged whether further processing is needed. According to the marked defect data set, the automatic comparison tool is matched with the preset standard for matching analysis, and the final quality judgment result is obtained.

[0028] Specifically, for the secondary assessment of weld surface quality and dynamic monitoring of grinding effects, a high-precision laser scanner is first used to collect three-dimensional data of the weld surface to obtain surface roughness data. Assume that the measured average roughness value is Ra 3.2 microns with a standard deviation of 0.5 microns. Subsequently, the surface quality monitoring mechanism is used to compare the collected data with the preset quality standard, which requires a roughness Ra of less than 2.8 microns. Through algorithm calculation, the current roughness value exceeds the standard value by 0.4 microns. The system automatically triggers the quality deviation analysis module, analyzing that the cause of the deviation may be improper setting of the grinding equipment parameters, such as a grinding wheel speed of 800 rpm, which is lower than the recommended 1000 rpm. Next, the system calls the historical data comparison algorithm and finds that the roughness was Ra 2.6 microns when the previous grinding speed was 950 rpm, which meets the standard. Therefore, it is inferred that insufficient speed is the main problem. Furthermore, the system automatically adjusts the grinding parameters, increases the rotation speed to 1000 rpm, and scans the surface again after the adjustment through the real-time monitoring module. Assuming that the newly measured roughness is Ra 2.7 microns, the standard deviation is reduced to 0.3 microns, which meets the preset standards. Finally, the system generates an evaluation report, recording the initial value of 3.2 microns, the adjusted value of 2.7 microns, and the parameter change process, forming a complete data chain to ensure the accuracy and traceability of quality assessment. At the same time, the system associates the results with the weld strength detection business. If the roughness meets the standard, the strength test module is automatically triggered to ensure the continuity of the subsequent process flow. Test data such as tensile strength must reach 500MPa or above before entering the next link. Through the above information technology means, the entire process is automated, with rigorous logic and interconnected links.

[0029] Step S108: If the polishing effect does not meet the preset quality standard, the polishing path optimization plan is readjusted, new polishing parameters and trajectory data are obtained, and the polishing operation is executed cyclically.

[0030] Through real-time monitoring of the polishing effect, the evaluation data of the current polishing effect is obtained to determine whether it meets the preset quality standards. If the polishing effect does not meet the preset quality standards, the evaluation data is analyzed, and a new polishing plan is generated using a path optimization tool to obtain adjusted trajectory data. According to the adjusted trajectory data, the polishing parameters are updated and the new parameter configuration information is determined. According to the new parameter configuration information, the polishing equipment is driven to perform a cyclic operation to obtain updated polishing effect data. The updated polishing effect data is compared with the preset quality standards to determine whether it meets the requirements. If the comparison result still does not meet the requirements, the polishing process is adjusted through the data update mechanism in combination with the effect evaluation information to obtain an optimized execution plan. According to the optimized execution plan, the polishing process is reconfigured and the above steps are executed cyclically until the preset quality standards are met.

[0031] Specifically, in the case where the polishing effect of the weld does not meet the preset quality standard, the system automatically starts the path optimization and parameter adjustment process, achieving fully automated processing through information technology. First, the system analyzes the current polishing trajectory using the built-in path planning algorithm. Assuming that the initial trajectory coverage is 85%, which is lower than the ideal 95%, resulting in uneven polishing in some areas, the system recalculates the path through the trajectory optimization module, generates new trajectory data, and increases the coverage to 96% and adjusts the polishing angle from 30 degrees to 45 degrees to enhance the contact effect. Subsequently, the system analyzes the parameter records in the historical database and finds that the polishing pressure in the previous successful case was 5.5 Newton, while the current pressure is only 4.2 Newton, with a deviation of 1.3 Newton, which may lead to insufficient polishing depth. Therefore, the pressure parameter is automatically adjusted to 5.8 Newton to compensate for potential errors. Next, the system predicts the effect of the adjusted parameters through simulation algorithms, calculates that the new trajectory and parameter combination may reduce the surface roughness from the initial 4.1 microns to 2.9 microns, close to the standard value of 3.0 microns or below. Further, the system transmits the new parameters and trajectory data to the device control module, executes the polishing operation in a loop, and collects vibration data during the execution through a real-time feedback mechanism. Assuming that the vibration frequency is stable within 10 Hz, indicating that the device is running normally, if it exceeds the range, the pressure is automatically adjusted to 5.6 Newton to maintain stability. Finally, the system associates the polished data with the subsequent surface defect detection business. If the roughness meets the standard, the defect scanning module is automatically triggered to ensure that the weld surface is free of micro-cracks, with a scanning resolution of 0.1 microns. This forms a complete process chain, ensuring that each step is automatically processed by the system, with a logical and traceable data.

[0032] Step S109, after the polishing effect meets the preset quality standard, record the final surface quality monitoring data and defect recognition accuracy information, generate a complete detection and polishing report and store it in the system database.

[0033] Through real-time collection of polishing effect data, preliminary evaluation results of surface quality are obtained to determine whether they meet the preset quality standard. If the preliminary evaluation results of surface quality meet the preset quality standard, key indicators in the monitoring data are extracted to obtain final data for subsequent analysis. According to the final data, a pre-established defect recognition model is used to classify surface defects using a support vector machine algorithm to determine the defect type and distribution. If the defect type and distribution are within an acceptable range, further analyze the accuracy information to obtain the accuracy data of defect recognition to determine whether it meets the preset threshold. By comparing the accuracy data, the comprehensive evaluation results of quality monitoring are obtained to determine the final state of the surface quality. According to the comprehensive evaluation results, the structured content of the detection report is generated and stored in the system database to complete the data archiving.

[0034] Specifically, after the polishing effect reaches the preset quality standard, the system automatically triggers the recording process of surface quality monitoring data. It uses a high-precision optical scanner to conduct a comprehensive inspection of the workpiece surface and obtain surface roughness data. For example, the measured average roughness value is Ra 0.8 microns, which meets the preset standard range of Ra 1.0 microns. At the same time, the surface flatness deviation is recorded as 0.02 mm, which is lower than the allowable deviation of 0.05 mm. Subsequently, the system uses a deep learning-based defect recognition algorithm to analyze the scanned image and adopts a convolutional neural network model (CNN) to classify and locate tiny surface defects such as scratches and pits. The recognition accuracy reaches 98.5%, of which the false alarm rate of scratch detection is 1.2%, and the missed alarm rate is 0.3%, both meeting the preset accuracy requirements (false alarm rate less than 1.5%, missed alarm rate less than 0.5%). Next, the system automatically generates a complete inspection and polishing report, including surface quality data, defect identification results, and image evidence. The report also includes key parameters in the polishing process, such as a polishing pressure of 5.5 Newtons and a rotation speed of 3000 rpm. The quality data is trend analyzed through the data analysis module, and the correlation coefficient between surface roughness and polishing parameters is calculated to be 0.85, indicating that parameter adjustment has a significant impact on quality, providing a basis for subsequent optimization. Finally, the report is stored in an encrypted format in the system database, using the SHA-256 algorithm to ensure data security. At the same time, a unique identification code (such as ID: 20231015-001) is generated and associated with the workpiece batch for easy traceability. The database is also connected to the production management system to automatically update the workpiece status to "Quality Inspection Passed", realizing full process automation from detection to storage, ensuring data integrity and business continuity.

[0035] It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any appropriate manner without contradiction. In order to avoid unnecessary repetition, the present invention will not further describe various possible combinations. In addition, the various different embodiments of the present invention can also be arbitrarily combined, as long as they do not violate the concept of the present invention, they should also be regarded as the content disclosed by the present invention.

Claims

1. A stainless steel rotating body weld inspection and grinding method, characterized in that: The method comprises: The industrial camera equipment is used to collect real-time images of the weld surface of the stainless steel rotating body, and the collected raw image data is preliminarily stored and format converted to obtain weld surface image data for subsequent processing; Image processing algorithms are used to perform denoising and edge enhancement on weld surface image data, and feature extraction is performed on the processed images to determine the distribution of potential defect areas on the weld surface. Based on the distribution of potential defects, a pre-established defect classification model is used to analyze the defect type and severity to determine whether the defects have reached the preset threshold standard that triggers a production stoppage. If the severity of the defect reaches the preset threshold, a defect trigger instruction is generated, and a shutdown signal is sent to the relevant equipment through the shutdown control mechanism to obtain shutdown execution status feedback; Based on the feedback from the production stoppage execution status, an initial plan for grinding path optimization is generated based on the specific location and characteristics of the defect area, and the movement trajectory and parameter settings of the grinding tool are determined; Through the polishing path optimization solution, the defect area is accurately located and polished, and the surface image data of the polishing process is collected in real time to obtain dynamic monitoring results of the polishing effect; Based on the dynamic monitoring results of the grinding effect, the surface quality monitoring mechanism is used to conduct a secondary assessment of the weld surface quality to determine whether the grinding meets the preset quality standards; If the polishing effect does not meet the preset quality standards, the polishing path optimization plan is readjusted, new polishing parameters and trajectory data are obtained, and the polishing operation is executed cyclically; After the polishing effect meets the preset quality standards, the final surface quality monitoring data and defect identification accuracy information are recorded, and a complete inspection and polishing report is generated and stored in the system database.

2. A stainless steel rotating body weld inspection and polishing method according to claim 1, characterized in that: The method of collecting real-time images of the weld surface of the stainless steel rotating body by industrial camera equipment, performing preliminary storage and format conversion on the collected original image data, and obtaining weld surface image data for subsequent processing includes: The raw image data is obtained by performing real-time image acquisition on the weld surface of the stainless steel rotating body at a preset acquisition frequency using an industrial camera; If the format of the original image data does not meet the preset processing requirements, converting the original image data into a target format using a data conversion tool to obtain the first image data; Based on storage capacity limitations, preliminarily storing the first image data using a compression algorithm to obtain the compressed image data; The compressed image data is acquired and the image quality is adjusted using an image enhancement algorithm to obtain the weld surface image data.

3. A stainless steel rotating body weld inspection and grinding method according to claim 1, characterized in that: The image processing algorithm is used to perform denoising and edge enhancement on the weld surface image data, and feature extraction is performed on the processed image to determine the distribution of potential defect areas on the weld surface, including: Performing preliminary processing on the weld surface image data using image processing technology, and obtaining first image data through a denoising operation; Applying edge enhancement technology to the first image data to obtain second image data; Extracting key feature information based on the second image data and determining the distribution of significant features on the weld surface; By analyzing the distribution of significant features, if the feature distribution deviates from the preset threshold range, it is determined that there is a potential defect area and preliminary defect location information is obtained; Based on the preliminary defect location information and the spatial distribution characteristics of the weld surface, a convolutional neural network in the deep learning model is used for refined classification to obtain a specific classification of defect types. According to the specific classification of defect types, the defect area is spatially mapped to determine the precise distribution range of the defect area; By integrating data on the precise distribution range of defect areas, comprehensive distribution information of weld surface defects is generated.

4. A stainless steel rotating body weld inspection and grinding method according to claim 1, characterized in that: The aforementioned method uses a pre-established defect classification model to analyze the defect type and severity based on the potential defect area distribution to determine whether the defect reaches the preset threshold standard for triggering a production stoppage, including: Obtain potential defect area distribution data, standardize and clean the data through the preprocessing module to obtain a structured defect distribution data set; Use the pre-established random forest model to classify the structured defect distribution dataset, analyze the defect types, and obtain defect type labels; Use the support vector machine model to evaluate the severity of defect type labels and generate defect severity scores; If the defect severity score exceeds the preset threshold, the probability of production stoppage is predicted through the logistic regression model to obtain the possibility of production stoppage triggering; Based on the possibility of production stoppage triggering and combined with the distribution feature analysis module, high-risk defect areas are extracted and a priority list of defect areas is generated; Through the defect area priority list, the automated decision module is called to determine whether the production suspension criteria are met and generate a production suspension decision instruction; According to the production suspension decision instruction, the production monitoring database is updated, the defect handling status is recorded, and the basis for production adjustment is obtained.

5. The stainless steel rotating body weld inspection and grinding method according to claim 1, characterized in that: If the severity of the defect reaches a preset threshold, a defect trigger instruction is generated, a shutdown signal is sent to the relevant equipment through the shutdown control mechanism, and shutdown execution status feedback is obtained, including: Monitor the severity of defects in real time through the defect detection system, compare it with the preset threshold standards, and determine whether the trigger conditions are met; If the severity of the defect exceeds the preset threshold, the command generation logic is used to construct a defect trigger command, determine the command priority and target device; Based on the output of the instruction generation logic, a production stop control mechanism is used to send a production stop signal to the relevant equipment and obtain the transmission status of the signal transmission path; After obtaining the transmission status from the signal transmission path, the system determines whether the production stop signal has been successfully delivered based on the reception status of the relevant equipment signal and obtains the equipment response information; If the device response information indicates that the stop signal was not successfully delivered, the stop signal is regenerated through the feedback processing mechanism and the adjustment plan of the signal transmission path is determined; According to the adjusted signal transmission path, the shutdown signal is resent to obtain the shutdown execution status and determine whether the equipment operation status meets expectations; If the equipment operating status is as expected, the shutdown execution status is recorded through status feedback information to generate complete execution log data.

6. A stainless steel rotating body weld inspection and grinding method according to claim 1, characterized in that: Based on the feedback of the production suspension execution status, an initial plan for grinding path optimization is generated according to the specific location and characteristics of the defect area, and the movement trajectory and parameter settings of the grinding tool are determined, including: Obtain relevant data on defect areas from the execution feedback of the production suspension status, perform preliminary analysis on the feedback information, extract descriptive information of specific locations and regional characteristics, and obtain preliminary mapping results of defect distribution; Based on the preliminary mapping results of the defect distribution, classification processing is carried out according to the specific location and regional characteristics, and the severity of the defects is graded using preset thresholds to determine the priority order for each area; By prioritizing the order and combining the data of specific locations, an initial planning scheme for the polishing path is generated. The A-star algorithm in the path planning algorithm is used to perform preliminary calculations on the movement trajectory to obtain a preliminary layout of the path. Based on the preliminary layout of the path, the parameters of the grinding tool are adjusted according to the regional characteristics and the severity of the defects. If the parameter configuration does not match the preset threshold range, the movement trajectory is locally corrected to determine the optimized path plan; The optimized path plan is combined with real-time data from tool control to dynamically adjust the movement trajectory. If abnormal fluctuations are detected in the real-time data, the parameter configuration is recalibrated to obtain the final trajectory data. Based on the final trajectory data, the integrity of the path plan is verified, and historical records in the execution feedback are obtained for comparison. If there are any deviations, the path plan is fine-tuned to determine the final execution plan. Through the final execution plan, control instructions for the grinding tool are generated, the movement trajectory and parameter configuration are monitored in real time, and the status data during the execution process is obtained to determine whether it meets the preset standards.

7. The stainless steel rotating body weld inspection and grinding method according to claim 1, characterized in that: The polishing path optimization solution accurately locates and polishes the defective area, collects surface image data in real time during the polishing process, and obtains dynamic monitoring results of the polishing effect, including: The surface image data of the grinding process is acquired in real time by the camera, and the defect features are extracted using image segmentation technology to determine the coordinate position of the defect area; According to the coordinate position of the defect area, the A-star algorithm is used to generate the shortest grinding path and determine the movement trajectory of the grinding equipment; By comparing the surface image data with a preset standard image, if the pixel difference between the surface image data and the preset standard image is less than a preset threshold, it is determined that the polishing effect is qualified and the deviation data is obtained; The force parameters and path trajectory of the grinding equipment are adjusted according to the deviation data. If the deviation data exceeds a preset range, the path is recalculated and the force parameters are updated, and the adjusted effect data is stored.

8. The stainless steel rotating body weld inspection and grinding method according to claim 1, characterized in that: Based on the dynamic monitoring results of the grinding effect, the surface quality monitoring mechanism is used to conduct a secondary assessment of the weld surface quality to determine whether the grinding meets the preset quality standards, including: The grinding effect data of the weld surface is obtained through real-time monitoring, and sensor technology is used to collect surface quality related information to obtain a preliminary monitoring data set; Based on the preliminary monitoring data set, feature extraction is performed on the weld surface, and the pre-established surface quality feature library is used to determine the initial surface quality assessment results; If the initial assessment results do not meet the preset standards, the monitoring mechanism will conduct an in-depth analysis of the grinding effect on the weld surface to obtain the specific distribution information of the abnormal area and determine whether there are local defects; Based on the distribution information of abnormal areas, image processing technology is used to scan the weld surface in detail to obtain high-resolution surface quality image data; Through high-resolution surface quality image data, the support vector machine algorithm is used to classify the types of defects on the weld surface and determine the specific category and severity of the defects; If the defect category and severity exceed the preset standards, data is marked for the defect area, and the marked defect data set is obtained to determine whether further processing is required; Based on the marked defect data set, an automated comparison tool is used to perform matching analysis with the preset standards to obtain the final quality judgment result.

9. The stainless steel rotating body weld inspection and grinding method according to claim 1, characterized in that: If the polishing effect does not meet the preset quality standard, the polishing path optimization plan is readjusted, new polishing parameters and trajectory data are obtained, and the polishing operation is cyclically executed, including: Through real-time monitoring of the polishing effect, the evaluation data of the current polishing effect is obtained to determine whether it meets the preset quality standards; If the polishing effect does not meet the preset quality standards, the evaluation data is analyzed and a new polishing plan is generated using a path optimization tool to obtain adjusted trajectory data; According to the adjusted trajectory data, the polishing parameters are updated and the new parameter configuration information is determined; Drive the grinding equipment to execute cyclic operation through the new parameter configuration information to obtain updated grinding effect data; Compare the updated polishing effect data with the preset quality standards to determine whether they meet the requirements; If the comparison results still do not meet the requirements, the polishing process will be adjusted through the data update mechanism and combined with the effect evaluation information to obtain an optimized implementation plan; According to the optimized execution plan, the polishing process is reconfigured and the above steps are repeated until the preset quality standards are met.

10. The stainless steel rotating body weld inspection and grinding method according to claim 1, characterized in that: After the polishing effect meets the preset quality standards, the final surface quality monitoring data and defect recognition accuracy information are recorded, and a complete inspection and polishing report is generated and stored in the system database, including: By collecting real-time data on the polishing effect, we can obtain preliminary evaluation results of the surface quality and determine whether it meets the preset quality standards; If the preliminary assessment results of the surface quality meet the preset quality standards, key indicators in the monitoring data are extracted to obtain the final data for subsequent analysis; Based on the final data, a pre-established defect recognition model is used to classify surface defects using the support vector machine algorithm to determine the defect type and distribution; If the defect type and distribution are within an acceptable range, the accuracy information is further analyzed to obtain the accuracy data of the defect identification and determine whether it meets the preset threshold; By comparing the accuracy data, we can obtain the comprehensive evaluation results of quality control and determine the final status of surface quality; Based on the comprehensive evaluation results, the structured content of the test report is generated and stored in the system database to complete data archiving.

Images