A method for attributing edge errors of freeform parts

Abstract

The present application relates to the technical field of mould, and particularly relates to a cutting-free part edge error attribution method. The cutting-free part edge error attribution method comprises the following steps: S10, system calibration and model import; S20, synchronous image acquisition and edge extraction; S30, coordinate unification and deviation calculation; S40, deviation root source judgment and positioning compensation; S50, data storage and mould state analysis. Compared with the prior art, the cutting-free part edge error attribution method can detect the part edge size on line and in real time, can trace the deviation root source, and realizes the intelligent method of feedforward compensation and mould state early warning.

Description

Technical Field

[0001] This invention relates to the field of mold technology, specifically a method for attributing edge errors in cut-free parts. Background Technology

[0002] To meet the increasingly stringent requirements of the automotive industry for lightweighting, safety, and cost control, hot stamping with no trimming has become a mainstream development trend. Its core concept is to precisely calculate the deformation of parts during hot stamping, directly ensuring that the stamped parts meet the dimensional and quality requirements of the final product, thus eliminating the need for subsequent expensive laser cutting or mechanical trimming processes. This not only significantly reduces production energy consumption, shortens the process flow, and improves material utilization, but also avoids weakening or deformation of parts caused by secondary processing.

[0003] However, achieving stable and reliable edge-free production faces a series of severe challenges, the most critical of which lies in the precision control of part edge dimensions. The main influencing factors can be summarized into the following three levels: 1. Random errors in blank positioning and process transfer: After heating, the sheet metal is in a red-hot, softened state. During the entire process of being picked up by the loading robot, conveyed by the conveyor, and finally placed in the lower mold, random positional deviations are easily generated due to mechanical vibration, thermal deformation, the fit clearance of positioning pins / holes, and slight drifts of the vision or mechanical positioning system. This initial deviation will directly map to the edge position of the formed part, resulting in hole offset and out-of-tolerance contours, seriously affecting the assemblability and interchangeability of the part.

[0004] 2. Gradual Deterioration of Die Working Condition: Hot stamping dies operate under extreme conditions of high temperature, high pressure, and cyclic thermo-mechanical fatigue. Their critical working surfaces inevitably experience wear and minor deformation. This damage accumulates slowly and gradually, resulting in systematic dimensional shifts at the edges of the formed parts, such as gradually increasing hole diameter, gradually narrowing or widening contours, or increasing burr height. In traditional production methods, these changes are difficult to detect in real time.

[0005] 3. The impact of process parameter fluctuations: Fluctuations in process parameters such as sheet heating temperature, mold temperature, holding pressure and time, and cooling rate can also cause slight changes in the microstructure and shrinkage rate of the parts, thus having a cumulative effect on the stability of edge dimensions.

[0006] Currently, the control of part precision mainly relies on mold debugging before production and periodic offline sampling inspections. This method cannot respond in real time to random deviations in a single stamping process, nor can it detect progressive mold wear in a timely manner, resulting in a high risk of defective products and delayed mold maintenance, which may lead to batch quality accidents. Summary of the Invention

[0007] To overcome the shortcomings of the prior art, this invention provides a method for attributing edge errors without cutting parts. This method can detect the edge dimensions of parts online and in real time, trace the root cause of deviations, and realize intelligent methods for feedforward compensation and mold status early warning.

[0008] To achieve the above objectives, a method for attributing edge errors in parts without cutting is designed, characterized by the following steps: S10, System Calibration and Model Import: The internal and external parameters of multiple industrial cameras arranged at the inspection station are calibrated, and the transformation relationship from the image pixel coordinate system to the world coordinate system is established; at the same time, the CAD design model of the target part is imported into the system, and the ideal boundary feature data representing the part contour and hole position are extracted from it. S20, Synchronous Image Acquisition and Edge Extraction: After the part is transferred to the fixed inspection station, the industrial camera is activated to extract images of specific edge areas to be measured; the acquired images are preprocessed, and a sub-pixel precision edge detection algorithm is used to extract the sub-pixel coordinate sequence of the actual edge of the part; S30, Coordinate unification and deviation calculation: Using calibration parameters, the local edge coordinates extracted by each camera are uniformly transformed to the world coordinate system with the detection platform reference sphere as the origin; the actual contour coordinate data and ideal boundary feature data are spatially registered and compared, and the quantitative deviation data including at least the positional deviation of key feature points, the overall offset of the contour line and the contour matching degree are calculated. S40, Deviance Root Cause Judgment and Positioning Compensation: Analyze and quantify the spatial distribution pattern of deviation data; if the deviation pattern is determined to conform to the characteristics of systematic positioning error, then generate compensation control instructions for the upstream feeding or positioning mechanism based on the deviation vector, which are used to adaptively adjust the placement position or posture of the next blank to be processed. S50, Data Storage and Mold Status Analysis: Continuously stores part identification, quantitative deviation data, and executed compensation instructions corresponding to each production cycle; performs statistical analysis on deviation data of specific feature points based on time series to identify their long-term trend; if a monotonic trend of deviation at a specific location is found and exceeds the threshold, mold wear is determined and an early warning is issued.

[0009] In step S40, the systematic positioning error is characterized by the high consistency of the position deviation vectors of multiple discrete feature points in the translation direction and / or rotation angle; the upstream feeding or positioning mechanism includes at least one of a feeding robot, an adjustable blank positioning block, or an adjustable positioning pin in the mold.

[0010] In step S40, generating compensation control instructions includes performing digital filtering or weighted averaging on the systematic deviation vectors calculated in the current and previous production cycles to eliminate random noise, obtain a stable compensation amount, and then convert it into control instructions for the actuators.

[0011] In step S50, the time-series-based statistical analysis includes calculating the moving average of the deviation data of specific feature points or performing linear or nonlinear regression fitting; the wear warning threshold is dynamically set according to the mold design tolerance, part functional requirements or historical experience data.

[0012] It also includes step S60, process parameter correlation analysis: the quantitative deviation data is correlated with the process parameter data of the hot stamping production line. The process parameters include at least the sheet heating temperature, mold temperature, holding pressure and cooling water flow rate; the analysis is conducted to see if the relevant process parameters fluctuate abnormally when a specific deviation pattern occurs, so as to help determine the compound cause of the deviation.

[0013] In step S20, before or during image acquisition, a specific illumination source is projected onto the edge area of ​​the part to be measured in order to enhance edge contrast or obtain three-dimensional height information of the edge.

[0014] A machine vision-based online control system for edge dimensions of hot-stamped parts and a mold condition monitoring system, characterized in that it includes: The visual inspection platform includes at least four high-resolution industrial cameras arranged at inspection points to capture images of different positions on the parts, and a high-brightness LED light source array that works with the industrial cameras to provide uniform illumination to the edges of the parts. The core processing and computing unit, connected to the visual inspection platform via signal, includes an image processing module, a coordinate transformation and registration module, a deviation calculation and analysis module, and a compensation decision module. The image processing module performs image preprocessing and edge extraction. The coordinate transformation and registration module performs coordinate system and model comparison. The deviation calculation and analysis module calculates and quantifies the deviation and determines the deviation pattern. The compensation decision module generates positioning compensation commands. The control and execution unit is communicatively connected to the core processing and computing unit and the main control system of the hot stamping production line. It is used to receive the compensation instructions and drive the feeding robot or the blank positioning mechanism in the mold to perform compensation actions. The data management and early warning unit, connected to the core processing and computing unit, includes a database for storing all production process data, a data analysis module for performing edge deviation trend analysis of parts and model training, and a human-computer interaction interface for displaying real-time results, historical curves, and issuing early warning information.

[0015] The industrial camera in the vision inspection platform is a black-and-white camera or a color camera, and the frame rate is not lower than the production line cycle time requirement; the high-brightness LED light source array is a ring light source, a strip light source, or a coaxial light source, and the emission color and angle are adjusted to adapt to different surface conditions of parts.

[0016] The core processing and computing unit also includes a self-learning optimization module. Based on historical compensation data and the final part deviation results, the self-learning optimization module optimizes the filtering parameters, compensation coefficients, or deviation mode judgment logic in the compensation decision module through machine learning algorithms. In addition, it combines process parameters in the production process, including sheet heating temperature, mold temperature, holding pressure, and cooling water flow rate, to determine the impact of different parameter changes on the edge dimensions of the sheet.

[0017] The control and execution unit communicates with the main control system of the production line via industrial Ethernet or fieldbus; the blank positioning mechanism is a miniature in-mold blank baffle driven by a servo motor or an adjustable positioning pin that can be precisely positioned.

[0018] A hot stamping production line includes the machine vision-based online control system for edge dimensions of hot stamped parts and mold status monitoring system described in any one of the above claims.

[0019] A computer-readable storage medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the steps of any of the methods described above.

[0020] Compared with the prior art, the present invention provides a method for attributing edge errors of parts without cutting, which can detect the edge dimensions of parts online and in real time, trace the root cause of deviations, and realize intelligent methods for feedforward compensation and mold status early warning. Attached Figure Description

[0021] Figure 1 This is a flowchart of the method of the present invention.

[0022] Figure 2 This is a schematic diagram of the detection bracket of the present invention.

[0023] Figure 3 This is a flowchart detailing the software process and algorithm implementation of the present invention.

[0024] Figure 4 This is a schematic diagram showing how the edge dimension deviation of a part changes with the number of times the mold is used.

[0025] Figure 5 , Figure 6 This is a schematic diagram of the deviation area of ​​the part edge due to mold wear.

[0026] Figure 7 This is the overall flowchart of the closed-loop control of the edge of the hot-stamped non-cut parts of the present invention. Detailed Implementation

[0027] The present invention will now be further described with reference to the accompanying drawings.

[0028] like Figure 1 As shown, the present invention provides a method for attributing edge errors of parts without cutting, comprising the following steps: S10, System Calibration and Model Import: The internal and external parameters of multiple industrial cameras arranged at the inspection station are calibrated, and the transformation relationship from the image pixel coordinate system to the world coordinate system is established; at the same time, the CAD design model of the target part is imported into the system, and the ideal boundary feature data representing the part contour and hole position are extracted from it. S20, Synchronous Image Acquisition and Edge Extraction: After the part is transferred to the fixed inspection station, the industrial camera is activated to extract images of specific edge areas to be measured; the acquired images are preprocessed, and a sub-pixel precision edge detection algorithm is used to extract the sub-pixel coordinate sequence of the actual edge of the part; S30, Coordinate unification and deviation calculation: Using calibration parameters, the local edge coordinates extracted by each camera are uniformly transformed to the world coordinate system with the detection platform reference sphere as the origin; the actual contour coordinate data and ideal boundary feature data are spatially registered and compared, and the quantitative deviation data including at least the positional deviation of key feature points, the overall offset of the contour line and the contour matching degree are calculated. S40, Deviance Root Cause Judgment and Positioning Compensation: Analyze and quantify the spatial distribution pattern of deviation data; if the deviation pattern is determined to conform to the characteristics of systematic positioning error, then generate compensation control instructions for the upstream feeding or positioning mechanism based on the deviation vector, which are used to adaptively adjust the placement position or posture of the next blank to be processed. S50, Data Storage and Mold Status Analysis: Continuously stores part identification, quantitative deviation data, and executed compensation instructions corresponding to each production cycle; performs statistical analysis on deviation data of specific feature points based on time series to identify their long-term trend; if a monotonic trend of deviation at a specific location is found and exceeds the threshold, mold wear is determined and an early warning is issued.

[0029] In step S40, the systematic positioning error is characterized by the high consistency of the position deviation vectors of multiple discrete feature points in the translation direction and / or rotation angle; the upstream feeding or positioning mechanism includes at least one of a feeding robot, an adjustable blank positioning block, or an adjustable positioning pin in the mold.

[0030] In step S40, generating compensation control instructions includes performing digital filtering or weighted averaging on the systematic deviation vectors calculated in the current and previous production cycles to eliminate random noise, obtain a stable compensation amount, and then convert it into control instructions for the actuators.

[0031] In step S50, the time-series-based statistical analysis includes calculating the moving average of the deviation data of specific feature points or performing linear or nonlinear regression fitting; the wear warning threshold is dynamically set according to the mold design tolerance, part functional requirements or historical experience data.

[0032] It also includes step S60, process parameter correlation analysis: the quantitative deviation data is correlated with the process parameter data of the hot stamping production line. The process parameters include at least the sheet heating temperature, mold temperature, holding pressure and cooling water flow rate; the analysis is conducted to see if the relevant process parameters fluctuate abnormally when a specific deviation pattern occurs, so as to help determine the compound cause of the deviation.

[0033] In step S20, before or during image acquisition, a specific illumination source is projected onto the edge area of ​​the part to be measured in order to enhance edge contrast or obtain three-dimensional height information of the edge.

[0034] A machine vision-based online control system for edge dimensions of hot-stamped parts and a mold condition monitoring system, characterized in that it includes: The visual inspection platform includes at least four high-resolution industrial cameras arranged at inspection points to capture images of different positions on the parts, and a high-brightness LED light source array that works with the industrial cameras to provide uniform illumination to the edges of the parts. The core processing and computing unit, connected to the visual inspection platform via signal, includes an image processing module 201, a coordinate transformation and registration module, a deviation calculation and analysis module, and a compensation decision module. The image processing module performs image preprocessing and edge extraction. The coordinate transformation and registration module performs coordinate system and model comparison. The deviation calculation and analysis module calculates and quantifies the deviation and determines the deviation pattern. The compensation decision module generates positioning compensation commands. The control and execution unit is communicatively connected to the core processing and computing unit and the main control system of the hot stamping production line. It is used to receive the compensation instructions and drive the feeding robot or the blank positioning mechanism in the mold to perform compensation actions. The data management and early warning unit, connected to the core processing and computing unit, includes a database for storing all production process data, a data analysis module for performing edge deviation trend analysis of parts and model training, and a human-computer interaction interface for displaying real-time results, historical curves, and issuing early warning information.

[0035] The industrial camera 101 in the visual inspection platform 100 is a black and white camera or a color camera, and the frame rate is not lower than the production line cycle time requirement; the high-brightness LED light source array 102 is a ring light source, a strip light source or a coaxial light source, and the light emission color and angle are adjusted to adapt to different part surface conditions.

[0036] The core processing and computing unit also includes a self-learning optimization module. Based on historical compensation data and the final part deviation results, the self-learning optimization module optimizes the filtering parameters, compensation coefficients, or deviation mode judgment logic in the compensation decision module through machine learning algorithms. In addition, it combines process parameters in the production process, including sheet heating temperature, mold temperature, holding pressure, and cooling water flow rate, to determine the impact of different parameter changes on the edge dimensions of the sheet.

[0037] The control and execution unit communicates with the main control system of the production line via industrial Ethernet or fieldbus; the blank positioning mechanism is a miniature in-mold blank baffle driven by a servo motor or an adjustable positioning pin that can be precisely positioned.

[0038] A hot stamping production line includes the aforementioned machine vision-based online control system for the edge dimensions of hot stamped parts and mold condition monitoring system.

[0039] A computer-readable storage medium having a computer program stored thereon, characterized in that: when the program is executed by a processor, it performs the steps of any of the methods described above.

[0040] Embodiments of the present invention: I. System Hardware Configuration and Installation The system in this embodiment is integrated into an automated hot stamping production line, located after the stamping machine unloading robot and before the parts stacking station.

[0041] 1. Image acquisition unit 100: Reference Figure 2 A separate, stable steel structure inspection platform 101 is designed, equipped with precision positioning pins 102 that match the positioning holes of the parts, ensuring that the parts maintain a consistent posture and are free from wobbling after each placement by the robotic arm. Four 5-megapixel global shutter monochrome industrial cameras 103 are mounted on this platform. They are placed around the parts to extract the edge dimensions of parts near areas with severe mold wear. Each camera corresponds to a set of high-brightness blue LED strip light sources 104. The light sources illuminate the edges of the parts at a low angle, creating a clear light-dark boundary line due to the height difference. To accommodate the differences in inspection points for different parts, the position and number of the crossbeams 105 that mount the cameras and light sources can be freely adjusted.

[0042] 2. Core Processing Unit: It adopts a high-performance industrial computer (IPC) with a built-in image acquisition card and gigabit network card, and communicates with four cameras through the GigE Vision protocol.

[0043] 3. Control and Execution Unit: In this embodiment, the compensation execution mechanism consists of a feeding robot and at least two servo-driven adjustable positioning pins within the mold. The installation position and number of positioning pins are adjusted according to the actual structural design of the part, and their extension height can be precisely adjusted within a certain range at the micrometer level by a servo motor, used to fine-tune the initial position of the blank within the mold cavity.

[0044] 4. Data Management Unit: Configure one database server and one engineering workstation (including human-machine interface HMI).

[0045] II. Software Flow and Algorithm Implementation Details refer to Figure 3 The specific implementation steps are as follows: 1. First stage: Initialization and calibration (corresponding to step S10).

[0046] (1) Camera Calibration: Using a high-precision checkerboard calibration board, at least 15 images were captured at multiple positions and angles on the testing platform. The Zhang Zhengyou calibration method was adopted, and the intrinsic parameters (focal length f_x, f_y, principal points c_x, c_y, distortion coefficients k1, k2, p1, p2) and extrinsic parameters (rotation matrix R and translation vector T relative to the world coordinate system) of each camera were simultaneously solved using the built-in algorithm. After calibration, the fields of view of the four cameras were unified into a world coordinate system {W} with the testing platform reference sphere as the origin.

[0047] (2) Part Model Import and Feature Definition: Extract all non-cutting contour lines and sufficiently large functional hole edges from the STEP format CAD model of the part as "ideal boundaries". Define the feature points or feature segments to be detected manually or automatically in the software.

[0048] 2. Second stage: Online detection and processing (completed within one production cycle, corresponding to steps S20-S30).

[0049] (1) Triggering and acquisition: After the part arrives at the inspection station and stops shaking, the photoelectric sensor sends a signal, and the IPC synchronously triggers four cameras to take pictures.

[0050] (2) Image processing: ①Preprocessing: Gaussian filtering is applied to each image for noise reduction, followed by contrast stretching.

[0051] ② Edge extraction: The Canny edge detection algorithm is used to initially locate the edges, and then combined with the Steger sub-pixel algorithm (based on the Hessian matrix) to perform sub-pixel precise location of the initial edges and obtain the sub-pixel coordinates (u, v) of the edge points.

[0052] ③ Coordinate transformation and stitching: Using the extrinsic parameters [R|T] of each camera obtained from calibration, the sub-pixel edge points (u, v) in each image are back-projected into the world coordinate system {W} to obtain a 3D point cloud (X_w, Y_w, Z_w). Since the height of the part in the Z direction on the inspection table is basically fixed, it can be simplified to XY plane processing.

[0053] (3) Registration and comparison: ① Coarse registration: Use feature-based matching (a reference datum should be provided at the edge of the part on the test fixture) to initially align the actual point set with the ideal model.

[0054] ② Precise registration: The Iterative Closest Point (ICP) algorithm is used to minimize the distance between the actual point set and the boundary line of the ideal model to achieve precise registration.

[0055] (4) Deviation calculation: After registration, the deviation (ΔX_i, ΔY_i) between the actual position and the theoretical position of each defined feature point P_i is calculated. At the same time, along the ideal contour line, samples are taken at fixed intervals to calculate the normal distance from the actual point cloud to the contour line, which is used as a measure of contour offset.

[0056] 3. Third stage: Decision making and control (corresponding to step S40).

[0057] (1) Deviation pattern judgment: According to the tolerance requirements of the specific parts, if the average value μ_x of all ΔX_i is significantly non-zero (e.g., |μ_x| > 0.3mm), while the mean value μ_y of ΔY_i is close to zero, and the variance of ΔX_i at each point is very small, then it is judged as an overall translation in the X direction. If (ΔX_i, ΔY_i) shows an obvious rotational distribution around a certain center, then it is judged as a rotational deviation.

[0058] (2) Compensation decision and execution: For the X-direction translation deviation, the compensation decision module uses a first-order low-pass digital filter to process the μ_x values ​​of N=5 consecutive parts to obtain the smoothed compensation amount C_x.

[0059] C(x) = α·C(x-1) +(1-α)·μ(x) + K_I·Σμ(i) + K_Pest·P_est Where α is the historical weighting coefficient, controlling the smoothness of the compensation change; the larger α is, the slower but more stable the response; the smaller α is, the faster the response but the more prone to oscillation. The recommended value is 0.65-0.75. (1-α) is the current weighting coefficient, which determines the response intensity of the current measurement value. The larger (1-α) is, the more the system pays attention to the current deviation. K_I is the integral coefficient, used to accumulate historical deviations and eliminate steady-state errors. The larger K_I is, the faster the elimination of residual error, but the more prone to overshoot. The recommended value is 0.01-0.05. K_Pest is the deviation estimation coefficient, used for feedforward compensation based on the inherent process deviation. It can accelerate convergence and reduce the integral burden. The recommended value is 0.005-0.03. Σμ(i) is the cumulative sum of all historical deviations, and P_est is the estimated inherent process deviation value.

[0060] (3) Generate instruction: Send the instruction "Adjust mold positioning pin, X-axis compensation -C_x mm" to the stamping line PLC via OPC UA protocol. The PLC drives the servo motor to adjust the position of the positioning pin by -C_x mm before the next sheet metal placement cycle.

[0061] To address rotational deviation, it is necessary to calculate the deflection angle of the sheet metal and adjust the rotation angle when placing the sheet metal by using two locating pins.

[0062] 4. Fourth stage: Data analysis and early warning (corresponding to step S50).

[0063] The database stores a record for each part, which includes: part ID, timestamp, all ΔX_i, ΔY_i, calculated μ_x, μ_y, contour offset value, and the amount of compensation C_x performed.

[0064] (1) Trend Analysis: The data analysis module automatically runs the analysis script every morning. Taking the "ΔX of the lower left corner edge dimension deviation" of a certain part as an example, the script reads the historical data of the parts working on that day and calculates its 30-day moving average. Figure 4 As shown, the total deviation of the edge size in this region was initially around -0.05 mm, and then began to show a stable negative growth trend on its own.

[0065] (2) Warning: When the moving average exceeds the first warning line of -0.15mm (approximately 30% of the tolerance zone), the system marks the hole location as "concerned" on the HMI and sends an email to the process engineer. When it exceeds the second warning line of -0.25mm (approximately 50% of the tolerance zone), the system issues a "mold maintenance" alarm, indicating that the rounded corners or cutting edges of the mold forming surface near the feature may need to be checked or repaired.

[0066] like Figure 5 , Figure 6The image shows a sheet metal model of a hot-stamped part. Taking the upper left corner edge as an example, the red solid line represents the ideal boundary of the part, and the black line represents the upper limit of the dimensional tolerance of this area. It can be seen that as the fillet near the mold detection area wears down, while the edges of the rest of the part approach the ideal boundary, the edge of the part in the detection area gradually shifts outward. When its edge contour line approaches the black line, the system will issue a "mold maintenance" alarm.

[0067] III. Extended Implementation Methods 1. Process Parameter Correlation: The system simultaneously collects data such as the temperature of the heating furnace zone and the inlet temperature of the mold cooling water via OPC UA. When a batch of parts exhibits abnormal contour expansion, and the mold cooling water temperature for that batch is recorded to be too high, the system can prompt "The dimensional abnormality may be related to insufficient cooling; it is recommended to stop the machine for inspection," assisting engineers in root cause analysis.

[0068] 2. Self-learning optimization: The self-learning optimization module continuously collects the "compensation instruction C_x" and the "residual deviation μ_x_residual actually measured in the next part after compensation". Through the composite control model, it continuously calibrates the compensation coefficients α, K_I and K_Pest (i.e., C(x) = α·C(x-1) +(1-α)·μ(x) + K_I·Σμ(i) + K_Pest·P_est) to optimize the compensation efficiency and anti-interference ability, thereby achieving self-improvement of system performance.

Claims

1. A method for attributing edge errors in non-cut parts, characterized in that, Includes the following steps: S10, System Calibration and Model Import: The internal and external parameters of multiple industrial cameras arranged at the inspection station are calibrated, and the transformation relationship from the image pixel coordinate system to the world coordinate system is established; at the same time, the CAD design model of the target part is imported into the system, and the ideal boundary feature data representing the part contour and hole position are extracted from it. S20, Synchronous Image Acquisition and Edge Extraction: After the part is transferred to the fixed inspection station, the industrial camera is activated to extract images of specific edge areas to be measured; the acquired images are preprocessed, and a sub-pixel precision edge detection algorithm is used to extract the sub-pixel coordinate sequence of the actual edge of the part; S30, Coordinate unification and deviation calculation: Using calibration parameters, the local edge coordinates extracted by each camera are uniformly transformed to the world coordinate system with the detection platform reference sphere as the origin; the actual contour coordinate data and ideal boundary feature data are spatially registered and compared, and the quantitative deviation data including at least the positional deviation of key feature points, the overall offset of the contour line and the contour matching degree are calculated. S40, Deviance Root Cause Judgment and Positioning Compensation: Analyze and quantify the spatial distribution pattern of deviation data; if the deviation pattern is determined to conform to the characteristics of systematic positioning error, then generate compensation control instructions for the upstream feeding or positioning mechanism based on the deviation vector, which are used to adaptively adjust the placement position or posture of the next blank to be processed. S50, Data Storage and Mold Status Analysis: Continuously stores part identification, quantitative deviation data, and executed compensation instructions corresponding to each production cycle; performs statistical analysis on deviation data of specific feature points based on time series to identify their long-term trends; If a deviation at a specific location is found to exhibit a monotonous changing trend and exceeds a threshold, mold wear is determined and an early warning is issued.

2. The method for attributing edge errors of non-cut parts according to claim 1, characterized in that: In step S40, the systematic positioning error is characterized by the high consistency of the position deviation vectors of multiple discrete feature points in the translation direction and / or rotation angle; the upstream feeding or positioning mechanism includes at least one of a feeding robot, an adjustable blank positioning block, or an adjustable positioning pin in the mold.

3. The method for attributing edge errors of non-cut parts according to claim 1, characterized in that: In step S40, generating compensation control instructions includes performing digital filtering or weighted averaging on the systematic deviation vectors calculated in the current and previous production cycles to eliminate random noise, obtain a stable compensation amount, and then convert it into control instructions for the actuators.

4. The method for attributing edge errors of non-cut parts according to claim 1, characterized in that: In step S50, the time-series-based statistical analysis includes calculating the moving average of the deviation data of specific feature points or performing linear or nonlinear regression fitting; the wear warning threshold is dynamically set according to the mold design tolerance, part functional requirements or historical experience data.

5. The method for attributing edge errors of non-cut parts according to claim 1, characterized in that: It also includes step S60, process parameter correlation analysis: the quantitative deviation data is correlated with the process parameter data of the hot stamping production line, and the process parameters include at least the sheet heating temperature, mold temperature, holding pressure and cooling water flow rate; Analyze whether relevant process parameters fluctuate abnormally when a specific deviation pattern occurs, in order to help determine the combined causes of the deviation.

6. The method for attributing edge errors of non-cut parts according to claim 1, characterized in that: In step S20, before or during image acquisition, a specific illumination source is projected onto the edge area of ​​the part to be measured in order to enhance edge contrast or obtain three-dimensional height information of the edge.

7. A machine vision-based online control system for edge dimensions of hot-stamped parts and a mold condition monitoring system for implementing the method described in any one of claims 1 to 6, characterized in that, include: The visual inspection platform includes at least four high-resolution industrial cameras arranged at inspection points to capture images of different positions on the parts, and a high-brightness LED light source array that works with the industrial cameras to provide uniform illumination to the edges of the parts. The core processing and computing unit, connected to the visual inspection platform via signal, includes an image processing module, a coordinate transformation and registration module, a deviation calculation and analysis module, and a compensation decision module. The image processing module performs image preprocessing and edge extraction. The coordinate transformation and registration module performs coordinate system and model comparison. The deviation calculation and analysis module calculates and quantifies the deviation and determines the deviation pattern. The compensation decision module generates positioning compensation commands. The control and execution unit is communicatively connected to the core processing and computing unit and the main control system of the hot stamping production line. It is used to receive the compensation instructions and drive the feeding robot or the blank positioning mechanism in the mold to perform compensation actions. The data management and early warning unit, connected to the core processing and computing unit, includes a database for storing all production process data, a data analysis module for performing edge deviation trend analysis of parts and model training, and a human-computer interaction interface for displaying real-time results, historical curves, and issuing early warning information.

8. The system for online control of edge dimensions of hot-stamped parts and monitoring of mold status based on machine vision, as described in claim 7, is characterized in that: The industrial camera in the vision inspection platform is a black-and-white camera or a color camera, and the frame rate is not lower than the production line cycle time requirement; the high-brightness LED light source array is a ring light source, a strip light source, or a coaxial light source, and the emission color and angle are adjusted to adapt to different surface conditions of parts.

9. The system for online control of edge dimensions of hot-stamped parts and monitoring of mold status based on machine vision, as described in claim 7, is characterized in that: The core processing and computing unit also includes a self-learning optimization module. Based on historical compensation data and the final part deviation results, the self-learning optimization module optimizes the filtering parameters, compensation coefficients, or deviation mode judgment logic in the compensation decision module through machine learning algorithms. In addition, it combines process parameters in the production process, including sheet heating temperature, mold temperature, holding pressure, and cooling water flow rate, to determine the impact of different parameter changes on the edge dimensions of the sheet.

10. The system for online control of edge dimensions of hot-stamped parts and monitoring of mold status based on machine vision, as described in claim 7, is characterized in that: The control and execution unit communicates with the main control system of the production line via industrial Ethernet or fieldbus; the blank positioning mechanism is a miniature in-mold blank baffle driven by a servo motor or an adjustable positioning pin that can be precisely positioned.

11. A hot stamping production line, characterized in that: The system includes a machine vision-based online control system for edge dimensions of hot-stamped parts and a mold condition monitoring system as described in any one of claims 7 to 10.

12. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, it implements the steps of the method as described in any one of claims 1 to 6.

Images