Blank forming defect risk perception and blank holder force self-feedback regulation system and method

Abstract

This invention discloses a sheet metal forming defect risk perception and blank holder force self-feedback control system and method, relating to the field of sheet metal forming technology. It aims to solve the technical problem in existing technologies where overall blank holder force control cannot suppress wrinkling, thinning, and cracking defects caused by local material flow imbalance in real time. The control system includes: a support; a punch mounted on the support; multiple hydraulic devices connected to the support; a blank holder ring divided into multiple blank holder sections, each section connected to the movable end of a hydraulic device; and a die slidably mounted on the support, positioned above the sheet metal and punch, movable towards the punch, working with the blank holder ring to press the outer edge of the sheet metal, and forming the sheet metal under the combined action of the die and the punch. This invention's sheet metal forming defect risk perception and blank holder force self-feedback control system completely overcomes the technical bottleneck of difficult defect control and unstable forming quality caused by the lag in response of traditional blank holder force adjustment methods by independently controlling the blank holder force in separate zones.

Description

Technical Field

[0001] This invention relates to the field of sheet metal forming technology, and in particular to a sheet metal forming defect risk perception and blank holder force self-feedback control system and method. Background Technology

[0002] Sheet metal drawing, as a typical plastic forming method, is widely used in advanced manufacturing fields such as automotive body panels and aerospace structural components. It is a key step in achieving lightweight, high-strength, and complex geometric shapes for parts. However, the forming accuracy of sheet metal is highly dependent on the synergistic effect of material properties, die structure, and process parameters. In actual production, it is affected by multiple coupled factors, easily leading to forming defects such as cracking and wrinkling, which seriously affects part quality and production efficiency. Especially in high-speed automated mass production lines, due to batch fluctuations in material properties and errors in die manufacturing and assembly, traditional process control methods based on preset process windows and manual experience adjustments are difficult to adapt to complex and changing actual working conditions, making it difficult to guarantee forming stability and part consistency.

[0003] Currently, common forming quality inspection methods in industrial settings mainly include: sampling inspection of formed sheet metal, forming limit measurement, and laser scanning of three-dimensional geometric contours for comparison. While these methods provide visualized forming quality data to some extent, they generally suffer from delayed response times, relying primarily on post-assessment and failing to achieve real-time control and process intervention. Traditional quality inspection methods have low levels of automation and intelligence, excessively relying on manual judgment and experience analysis, resulting in long debugging cycles, low efficiency, and a lack of stability and consistency. Therefore, if parts with defects such as cracking and necking are missed during inspection and flow into the next process, it will pose a significant threat to the overall vehicle's safety and quality assurance.

[0004] In light of this, driven by the rapid development of artificial intelligence, machine vision, and sensing technologies, visual inspection and image recognition technologies have been widely applied in manufacturing sites, alleviating to some extent the reliance on manual labor in traditional quality inspection methods. However, these technologies are currently still limited to the detection of appearance defects and have not yet achieved deep integration with the manufacturing process, especially in the prediction and prevention of forming defects at their source, where real-time control and process intervention are still not possible.

[0005] Therefore, to effectively solve the challenges of source control and process intervention for forming defects, the key lies in closely integrating image visual inspection technology with a self-feedback adjustment mechanism for process parameters. Typically, easily adjustable production parameters include blank holder force, oil application amount, and balance block. Adjusting the oil application amount often requires uniform cleaning of the entire batch of sheets and a revised oil application process, which is cumbersome, time-consuming, and currently impractical due to the difficulty in achieving real-time dynamic adjustment during forming. Adjusting the balance block controls the material flow by changing the die gap. However, on the production floor, adjusting the balance block requires rigorous process review and verification discussions to ensure the controllability of the adjustment process and the consistency of the final forming quality; therefore, it cannot be easily adjusted. Furthermore, balance block adjustment requires extremely high visual inspection accuracy, and the current accuracy errors in visual inspection can easily lead to inaccurate balance block adjustment, posing a significant application risk; therefore, adjusting the balance block is not the optimal solution. Among the many adjustable process parameters, blank holder force, as a key variable directly affecting the sheet material flow state and local stress distribution, has the most sensitive and direct control effect. Compared to variables that are difficult to adjust online, such as material properties or mold parameters, blank holder force offers excellent real-time controllability and is one of the most potentially adjustable process parameters in industrial settings. Especially in the drawing process of complex curved components, different regions exhibit significant differences in their response to blank holder force. If the force becomes unbalanced in areas with large curvature or corners, irreversible defects such as localized wrinkling or cracking can easily occur. After comprehensive evaluation, blank holder force, as a key parameter that can be adjusted in real time and significantly affects forming quality, is crucial for achieving real-time prediction and process intervention of sheet metal forming defects.

[0006] Based on the above analysis, there is an urgent need for an intelligent forming system that integrates image visual inspection technology with the ability to adjust the closed loop of the regional pressing force. This system can simultaneously achieve accurate identification, dynamic early warning, and real-time control of defects in the forming of sheet metal on site, thereby improving the stability of the forming process and the consistency of the products. Summary of the Invention

[0007] The present invention aims to solve or improve the technical problem in the prior art that the overall blank holder force control cannot suppress wrinkling, thinning and cracking defects caused by local material flow imbalance in real time.

[0008] The first aspect of the present invention is to provide a sheet metal forming defect risk perception and blank holder force self-feedback control system, comprising: a bracket; a punch mounted on the bracket; multiple hydraulic devices connected to the bracket; a blank holder ring divided into multiple blank holder sections, each blank holder section connected to the movable end of a hydraulic device, the sheet metal being disposed on the blank holder ring; and a die slidably mounted on the bracket, disposed above the sheet metal and the punch, movable toward the punch, working together with the blank holder ring to press the outer edge of the sheet metal, and, under the combined action of the die and the punch, processing the sheet metal into a component, wherein the component includes multiple detection areas, the detection areas being related to the blank holder force of the multiple blank holder sections. The affected sheet metal areas have a mapping relationship; an image acquisition component is configured to acquire images of the outer contour features of the component; an image analysis device is signal-connected to the image acquisition component and multiple hydraulic devices, and is configured to analyze the defect risk type in the outer contour feature image of each detection area, and generate a pressure compensation command for controlling the hydraulic device of the corresponding edge-pressing section based on the defect risk type and mapping relationship; after the current forming cycle ends and before the next forming cycle begins, the image analysis device sends the pressure compensation command to the corresponding hydraulic device to adjust the edge-pressing force of the corresponding edge-pressing section.

[0009] Optionally, in the above technical solution, the sheet metal forming defect risk perception and blank holder force self-feedback control system further includes: a detection platform, located on one side of the support; a robot arm, installed on the detection platform, which is used to grab components and transport them to the detection platform after the current forming cycle is completed; and an image acquisition component installed on the detection platform through an installation structure.

[0010] Optionally, in the above technical solution, the sheet metal forming defect risk perception and blank holder force self-feedback control system further includes: an electrically controlled slide rail structure, including a guide rail mounted on the mounting structure, a slider meshing with the guide rail, and a first servo motor driving the slider to move along the guide rail; an angle adjustment device mounted on the slider, including a horizontally rotatable gimbal and a second servo motor driving the gimbal's pitch angle; and an image acquisition component mounted on the gimbal.

[0011] Optionally, in the above technical solution, the image acquisition component is fixedly mounted on the mounting structure, and the mounting structure can be any structure used to mount the image acquisition component.

[0012] In the above technical solution, optionally, the image acquisition component includes: a housing mounted on a pan-tilt unit, with a light-transmitting glass at its bottom; a probe mounted inside the housing via a heat-insulating bracket, capable of acquiring images of the outer contour features of the sheet material through the light-transmitting glass; and a semiconductor temperature control module including: a thermoelectric cooler disposed on the inner wall of the housing; a temperature sensor disposed on the inner wall of the housing; and a PID controller connected to the thermoelectric cooler and the temperature sensor, wherein the PID controller is configured to activate the thermoelectric cooler to cool down when the temperature inside the housing is greater than a first preset temperature, and to activate the thermoelectric cooler to heat up when the temperature inside the housing is less than a second preset temperature.

[0013] Optionally, in the above technical solution, the sheet metal forming defect risk perception and blank holder force self-feedback control system further includes: a product specification database that stores the target coordinates and target attitude angles of the image acquisition components corresponding to different product numbers; a parameter mapping module that calls the data in the product specification database according to the input product number to generate displacement commands for the electronically controlled slide rail structure and angle commands for the gimbal; and a positioning execution unit that is connected to the parameter mapping module and, upon receiving the displacement and angle commands, controls the first and second servo motors to work to drive the slider to move to the target coordinates and simultaneously controls the gimbal to adjust to the target attitude angle.

[0014] In the above technical solution, optionally, the image acquisition component includes a camera and a lens, the camera includes an industrial camera, a security camera or a combination of the two, and the lens includes a fixed-focus lens or a zoom lens.

[0015] The second aspect of this invention provides a control method for a sheet metal forming defect risk perception and blank holder force self-feedback control system. The sheet metal forming defect risk perception and blank holder force self-feedback control system is any of the above-mentioned technical solutions. The control method includes: after the current forming cycle ends, acquiring the outer contour feature images of each detection area of ​​the component through an image acquisition component; analyzing the defect risk type in the outer contour feature images of each detection area through an image analysis device, and generating a pressure compensation command based on the defect risk type; within the time window from the end of the current forming cycle to the start of the next cycle, sending the pressure compensation command to the hydraulic device of the corresponding detection area via an industrial bus to adjust the blank holder force of the next forming cycle.

[0016] In the above technical solution, optionally, the step of analyzing the defect risk type in the outer contour feature image of each detection area using an image analysis device includes: constructing a finite element simulation model based on the property parameters of the sheet metal; conducting multi-condition simulation experiments based on the finite element simulation model, determining the first detection point on the sheet metal, and obtaining the take-up line length L of the first detection point under ideal forming conditions. 理 The length L of the take-up line at the critical state of wrinkling. 皱And the length L of the take-up line at the critical cracking state. 裂 Based on L 皱 and L 理 A safe critical size range for the take-up line was initially set for the first detection point; based on the critical size range of the take-up line initially determined by the finite element simulation model, and combined with on-site debugging, the simulation results were corrected and optimized, and finally a safe critical size range for the take-up line was set for the first detection point [L]. min L max ], where L min >L 皱 L max <L 裂 The image acquisition component is calibrated using a checkerboard calibration board, generating a transformation matrix from pixel coordinates to physical space coordinates. The position corresponding to the first detection point is determined in the outer contour feature image and denoted as the second detection point. The pixel distance from the second detection point to the outer edge of the sheet is calculated. This pixel distance is converted to physical space coordinates using the transformation matrix, and the actual take-up line length L of the second detection point is calculated. 实际 ; will L 实际 The critical size range of the receiving line [L] min L max Compare: If L 实际 < L min If L 实际 > L max If so, it is determined that there is a risk of cracking in the tested area.

[0017] In the above technical solution, optionally, the step of constructing a finite element simulation model based on the property parameters of the sheet metal specifically includes: constructing a three-dimensional geometric model in simulation software that includes the sheet metal, die, punch, and blank holder ring according to the geometric shape and structural dimensions of the sheet metal; setting a material constitutive model for the sheet metal; defining the contact relationship between the sheet metal and the die and punch, and setting the friction coefficient; applying the blank holder force parameter as a boundary condition to the contact surface between the blank holder ring and the sheet metal; and integrating the three-dimensional geometric model, material constitutive model, contact relationship, and boundary condition through a finite element solver to generate a finite element simulation model for simulating the forming process with different blank holder forces and friction coefficients.

[0018] Additional aspects and advantages of the invention will become apparent in the following description or may be learned by practice of the invention. Attached Figure Description

[0019] The above and / or additional aspects and advantages of embodiments of the present invention will become apparent and readily understood from the description of the embodiments in conjunction with the following drawings, wherein:

[0020] Figure 1 This shows one of the structural schematic diagrams of a sheet metal forming defect risk perception and blank holder force self-feedback control system according to an embodiment of the present invention;

[0021] Figure 2 This is a second schematic diagram of the structure of a sheet metal forming defect risk perception and blank holder force self-feedback control system according to an embodiment of the present invention;

[0022] Figure 3 A schematic diagram of the detection platform and electrically controlled slide rail structure according to an embodiment of the present invention is shown;

[0023] Figure 4 A schematic diagram of the structure of an image acquisition component according to an embodiment of the present invention is shown;

[0024] Figure 5 A flowchart of the control method of the sheet metal forming defect risk perception and blank holder force self-feedback control system according to an embodiment of the present invention is shown;

[0025] Figure 6 A schematic diagram of the drawing process according to an embodiment of the present invention is shown;

[0026] Figure 7 A Hockett-Sherby hardening model diagram according to an embodiment of the present invention is shown;

[0027] Figure 8 A diagram of the BBC2005 yield model according to an embodiment of the present invention is shown;

[0028] Figure 9 A forming limit curve diagram of an embodiment of the present invention is shown;

[0029] Figure 10 This diagram illustrates the change in the take-up line when the component of the present invention is at risk of cracking.

[0030] Figure 11 This diagram illustrates the changes in the take-up line when the component of the present invention is at risk of wrinkling.

[0031] Figure 12 A schematic diagram showing the changes in the take-up line when the component of the present invention is in a safe state is shown.

[0032] in, Figures 1 to 4 and Figure 6 The correspondence between the reference number and the component name is as follows:

[0033] 1. Bracket, 21. Punch, 22. Die, 23. Sheet metal, 24. Pressure ring, 32. First detection area, 34. Second detection area, 42. Hydraulic device, 44. Image acquisition component, 442. Housing, 444. Probe, 446. Semiconductor temperature control module, 448. Heat insulation bracket, 449. Image processing chip, 5. Slider, 6. Gimbal, 7. Nitrogen spring, 91. Detection platform, 92. Mounting structure. Detailed Implementation

[0034] To better understand the above aspects, features, and advantages of the embodiments of the present invention, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0035] like Figure 1 , Figure 2 and Figure 6 As shown, the first aspect of the present invention is to provide a sheet metal forming defect risk perception and blank holder force self-feedback control system. A punch 21 is mounted on a support 1; multiple hydraulic devices 42 are connected to the support 1; a blank holder ring 24 is divided into multiple blank holder sections, each blank holder section being connected to the movable end of a hydraulic device 42; a sheet metal 23 is disposed on the blank holder ring 24; a die 22 is slidably mounted on the support 1, positioned above the sheet metal 23 and the punch 21, and can move towards the punch 21, jointly pressing the outer edge of the sheet metal 23 with the blank holder ring 24, and processing the sheet metal 23 under the combined action of the die 21 to form a component. The component includes multiple detection areas, and the detection areas are connected to the multiple blank holder sections. The areas of the sheet metal affected by the blank holder force have a mapping relationship; an image acquisition component 44 is configured to acquire the outer contour feature image of the component; an image analysis device is signal-connected to the image acquisition component 44 and multiple hydraulic devices 42, and is configured to analyze the defect risk type in the outer contour feature image of each detection area, and generate a pressure compensation command for controlling the hydraulic device 42 of the corresponding blank holder section based on the defect risk type and mapping relationship; after the current forming cycle ends and before the next forming cycle begins, the image analysis device sends the pressure compensation command to the corresponding hydraulic device 42 to adjust the blank holder force of the corresponding blank holder section.

[0036] Specifically, during the entire forming process, the engagement of the punch 21, die 22, sheet metal 23, and pressure ring 24 is as follows:

[0037] Before the forming process begins, the hydraulic device 42 lifts the blank holder 24, whose height is higher than that of the punch 21. The sheet metal 23 is then placed above the blank holder 24, which is located at the bottom of the outer edge of the sheet metal 23. Subsequently, the die 22 moves downwards from above. When the die 22 contacts the sheet metal 23, the combined action of the die 22 above the sheet metal 23 and the blank holder 24 below it clamps the sheet metal 23, generating a draw bead. The die 22 then continues to move downwards, pulling the sheet metal 23 and the blank holder 24 downwards together, forcing the sheet metal 23 to undergo plastic deformation around the contour of the punch 21. During the downward movement of the die 22, the downward forming force of the die 22 is greater than the upward blank holder force provided by the hydraulic device 42; therefore, the blank holder force of the sheet metal 23 is equal to the hydraulic pressure provided by the hydraulic device 42.

[0038] The image acquisition component 44 is used to acquire the outer contour feature image of the component; the image analysis device is connected to the image acquisition component 44 and the hydraulic device 42 by signal, analyzes the defect risk type (such as wrinkling risk or cracking risk) in the outer contour feature image, generates a pressure compensation command, and sends the pressure compensation command to the corresponding hydraulic device 42 to adjust the blank holder force within the time window from the end of the current forming cycle to the start of the next cycle.

[0039] The sheet metal forming defect risk perception and blank holder force self-feedback control system of the present invention completely breaks through the technical bottleneck of difficult defect control and unstable forming quality caused by the lag response of traditional blank holder force adjustment methods by independently controlling the blank holder force in different zones.

[0040] Furthermore, this invention enables fully adaptive control throughout the entire process, eliminating the need for manual intervention and experience-based judgment. By detecting abnormal trends in material flow and outputting precise pressure compensation commands before macroscopic defects such as wrinkling or cracking occur, the system eliminates defects in their nascent stage, thereby achieving a fundamental shift from "passive remediation" to "proactive prevention," significantly improving the level of intelligent production and quality consistency.

[0041] The number and layout of the detection areas are directly determined by the number of independently controlled hydraulic devices 42, following a one-to-one mapping principle of "one area, one cylinder". Specifically, if the equipment of this application is equipped with n hydraulic devices, the edge pressing ring is synchronously divided into n independently adjustable edge pressing sections, and correspondingly, the sheet material 23 is also divided into n detection areas.

[0042] After the image acquisition component 44 acquires the outer contour feature image of the component, the image analysis device accurately determines the defect risk type of each detection area.

[0043] For example, if there is a risk of cracking in detection area A (manifested as insufficient material flow or excessive stretching), the system determines that the material flow in that area needs to be increased. The image analysis device then generates a pressure reduction command and sends it to the hydraulic device 42 controlling the blank holder section of detection area A, instructing it to reduce the blank holder force. The reduction in blank holder force reduces the resistance to material flow, thereby promoting more material to flow into the mold cavity and effectively alleviating the excessive stretching in that area.

[0044] This closed-loop control ensures that the next part avoids the risk of cracking in area A, while ensuring that other areas are not disturbed, ultimately resulting in a molded part with high forming accuracy and stable quality.

[0045] The detection area has a mapping relationship with the area of ​​the sheet material affected by the pressing force of multiple pressing sections. That is, each detection area is uniquely associated with a specific pressing section. When the image analysis device determines that there is a defect risk in a certain detection area, it can locate and generate a pressure compensation command for the hydraulic device to control the associated pressing section based on the mapping relationship. By independently adjusting the pressing force of the pressing section, the precise and directional control of the defect risk area can be achieved.

[0046] This mapping relationship is established through calibration during system initialization, ensuring a one-to-one correspondence between "where the imaging defect is, and where the pressure is adjusted," thereby completely solving the industry problem that traditional overall edge pressure adjustment cannot cope with local defects.

[0047] In the above technical solutions, such as Figure 3 As shown, the sheet metal forming defect risk perception and blank holder force self-feedback control system also includes: a detection platform 91, located on one side of the support 1, i.e., arranged in the second sequence of the production line; a robotic arm, installed on the detection platform 91, used to grasp components and transport them to the detection platform 91 after the current forming cycle ends; and an image acquisition component 44 installed on the detection platform 91 via the mounting structure 92.

[0048] In this technical solution, the sheet metal forming defect risk perception and blank holder force self-feedback control system also includes a detection platform 91, a robotic arm, and an installation structure 92. The image acquisition component 44 of this application is mounted on the detection platform 91 via the installation structure 92, rather than on the bracket 1. This allows the robotic arm to transfer the component to the independent detection platform 91 for image acquisition after the forming cycle is completed. This completely avoids the obstruction and interference of the blank holder ring 24 on the edge area of ​​the component in the traditional method, ensuring that the image acquisition component 44 can acquire clear and complete outer contour feature images of the entire component without blind spots. This significantly improves the accuracy and reliability of defect identification, providing a high-quality data foundation for the subsequent precise control of the blank holder force.

[0049] Of course, in another option, such as Figure 2 As shown, the image acquisition component 44 can also be mounted on the bracket 1, which simplifies the equipment.

[0050] Of course, the relative positional relationship between the component and the mold must not change during the process of the robot moving the sheet metal 23. After the robot moves the component to the standard position, the image acquisition component captures a panoramic image containing the entire component and its surrounding reference features.

[0051] The image analysis device can perform partitioning and defect identification according to the following process: The system first identifies pre-set visual reference points in the image. By calculating the offset between the actual pixel coordinates of the reference point in the image and the pre-stored standard pixel coordinates, the entire image is aligned and translated to ensure that the current component image and the pre-stored standard template in the system are in the same coordinate system. Subsequently, based on the pre-set virtual coordinates of each detection area, the system automatically extracts the sub-image corresponding to each area from the corrected panoramic image for subsequent independent analysis.

[0052] Furthermore, it should be noted that the complete processing of a finished part requires multiple steps. Among them, the core function of the first step is drawing and forming, which uses the cooperation of the first step punch 21 and die 22 to process the sheet metal 23 into the basic component shape.

[0053] The second stage is responsible for finishing or secondary drawing. The inspection platform 91 and the mounting structure 92 in this application are located in this second stage. The inspection platform 91 essentially acts as the punch of the second stage. Its surface structure is different from that of the punch 21 of the first stage. It is designed specifically to match the shape of the component after the first stage is formed, and is mainly used to support the component during the second stage processing.

[0054] The workflow is as follows: The robotic arm transfers the component formed in the first stage to the second stage and places it on the inspection platform 91 (i.e., the second-stage punch). Subsequently, the second-stage die descends and, together with the inspection platform 91, completes the subsequent processing steps.

[0055] Before the second processing, the image acquisition component 44 integrated on the mounting structure 92 acquires images of the component located on the detection platform 91 and transmits the data to the image analysis device to realize the self-feedback control of the blank holder force and provide process parameter optimization instructions for the next forming cycle.

[0056] In the above technical solutions, such as Figure 3As shown, the sheet metal forming defect risk perception and blank holder force self-feedback control system also includes: an electrically controlled slide rail structure, including a guide rail (not shown) mounted on the mounting structure 92, a slider 5 meshing with the guide rail, and a first servo motor (not shown) that drives the slider 5 to move along the guide rail; an angle adjustment device mounted on the slider 5, including a horizontally rotatable gimbal 6 and a second servo motor (not shown) that drives the gimbal 6 to tilt; and an image acquisition component 44 mounted on the gimbal 6.

[0057] In this technical solution, the sheet metal forming defect risk perception and blank holder force self-feedback control system also includes an electrically controlled slide rail structure and an angle adjustment device. The electrically controlled slide rail structure consists of a guide rail mounted on the mounting structure 92, a slider 5 meshing with the guide rail, and a first servo motor driving the slider 5 to move. The guide rail can be arranged longitudinally on the mounting structure 92 to achieve longitudinal movement of the slider 5, or it can be arranged laterally on the mounting structure 92 to achieve lateral movement of the slider 5, or both lateral and longitudinal guide rails can exist simultaneously to achieve both lateral and longitudinal movement of the slider 5. The angle adjustment device is fixedly mounted on the slider 5 and includes a 360° horizontally rotating gimbal 6 and a second servo motor driving the gimbal 6 to tilt (range ±30°). The image acquisition component 44 is mounted on the gimbal 6. Through the coordinated movement of the slider 5 translation and the gimbal 6 rotation / tilt, millimeter-level positioning accuracy of the detection position and precise adjustment of the shooting angle are achieved. This image detection structure can achieve full-area detection without blind spots.

[0058] In another parallel embodiment, the image acquisition component 44 is fixedly mounted on the mounting structure 92, which can be any structure for mounting the image acquisition component 44.

[0059] In the above technical solutions, optionally, such as Figure 4 As shown, the image acquisition component 44 includes: a housing 442, mounted on the gimbal 6, with a light-transmitting glass at its bottom; a probe 444, mounted inside the housing 442 via a heat-insulating bracket 448, capable of acquiring images of the outer contour features of the sheet metal 23 through the light-transmitting glass; and a semiconductor temperature control module 446, including: a semiconductor cooling chip, disposed on the inner wall of the housing 442; a temperature sensor, disposed on the inner wall of the housing 442; and a PID controller, connected to the semiconductor cooling chip and the temperature sensor, configured to activate the semiconductor cooling chip to cool down when the temperature inside the housing 442 is greater than a first preset temperature, and to activate the semiconductor cooling chip to heat up when the temperature inside the housing 442 is less than a second preset temperature.

[0060] In this technical solution, the image acquisition component 44 includes a housing 442 installed on the gimbal 6, the bottom of which is made of double-layer hollow anti-fog transparent glass (outer layer 5mm tempered hydrophobic glass + inner layer 3mm optical anti-reflective glass + 1.2mm argon hollow layer); the probe 444 is installed inside the housing 442 through a ceramic-based heat-insulating bracket 448, and the probe 444 faces the transparent glass to acquire the outer contour feature image of the sheet 23; the semiconductor cooling chip and temperature sensor of the semiconductor temperature control module 446 are attached to the inner wall of the housing 442, and the PID controller dynamically adjusts based on temperature feedback: when the temperature is greater than the first preset temperature (e.g., 32℃), cooling is started, and when the temperature is less than the second preset temperature (e.g., 18℃), reverse heating is started to maintain a constant temperature range inside the housing 442. This design incorporates double-layer anti-fog glass with an argon gas barrier to completely eliminate lens fogging. The heat insulation bracket 448 locks the operating temperature of the probe 444 at ≤38℃, preventing thermal damage to the probe 444 and significantly improving its service life. The semiconductor temperature control module 446 can maintain the temperature inside the housing 442, ensuring consistent measurement accuracy in workshop environments ranging from -10℃ to 45℃.

[0061] What needs to be understood, such as Figure 4 As shown, the main heating element of the image acquisition component 44 in this application is the image processing chip 449. The image processing chip 449 is connected to the probe 444. After the probe 444 acquires an image, the image is processed by the image processing chip 449 and then sent to the image analysis device. This application uses a heat insulation bracket 448 to isolate the image processing chip 449 from the probe 444, thereby improving the lifespan of the probe 444. That is, the heat insulation bracket 448 divides the outer shell 442 into a first cavity and a second cavity. The probe 444 is located in the first cavity, and the image processing chip 449 and the semiconductor temperature control module 446 are located in the second cavity.

[0062] Furthermore, the ambient temperature inside the workshop may also affect the lifespan of the probe 444 and the image clarity. Therefore, by adding the semiconductor temperature control module 446, the image clarity and the lifespan of the probe 444 are significantly improved.

[0063] Optionally, in the above technical solution, the sheet metal forming defect risk perception and blank holder force self-feedback control system also includes an attitude adjustment device. The attitude adjustment device includes: a product specification database that stores the target coordinates and target attitude angles of the image acquisition component 44 corresponding to different product numbers; a parameter mapping module that calls the parameters in the product specification database according to the input product number to generate displacement commands for the electronically controlled slide rail structure and angle commands for the gimbal 6; and a positioning execution unit that is connected to the parameter mapping module and, after receiving the displacement and angle commands, controls the first servo motor and the second servo motor to work, so as to drive the slider 5 to move to the target coordinates and simultaneously control the gimbal 6 to adjust to the target attitude angle.

[0064] In this technical solution, the attitude adjustment device includes a product specification database, a parameter mapping module, and a positioning execution unit. The product specification database stores the target coordinates (X, Y, Z) and target attitude angles (pitch angle θ, rotation angle φ) of the image acquisition component 44 corresponding to different product numbers. The parameter mapping module calls the data in the database according to the input product number and generates displacement commands for the electronically controlled slide rail structure and angle commands for the gimbal 6 in real time. The positioning execution unit drives the slider 5 to move to the target coordinates through the industrial bus-linked servo system, and synchronously controls the gimbal 6 to adjust to the target attitude angle, achieving fully automatic positioning. This device completely reconstructs the mold changeover process and breaks through the core bottleneck of flexible production of multiple varieties: the positioning time of the image acquisition component 44 is reduced from more than 25 minutes of manual debugging to within 30 seconds, the mold changeover efficiency of the production line is increased by more than 50 times, and it supports more than 20 high-frequency changeovers per hour; on the other hand, it can eliminate human error and reduce the image acquisition distortion rate to 0.1%, thereby reducing the technical dependence on operators.

[0065] In the above technical solution, optionally, the image acquisition component 44 includes a camera and a lens. The camera may include an industrial camera, a security camera, or a combination of both, and the lens may include a fixed-focus lens or a zoom lens. There are no fixed pairing restrictions between the camera and the lens; they can be combined arbitrarily.

[0066] In this technical solution, by configuring different types of cameras and lenses, different detection accuracy and field of view requirements can be flexibly adapted, thereby significantly improving the image acquisition effect.

[0067] In the above technical solution, optionally, the sheet metal forming defect risk perception and blank holder force self-feedback control system also includes a nitrogen spring 7, whose cylinder is integrated on the die 22, and whose end is connected to an elastic pressure plate. This nitrogen spring 7 is used to replace the traditional blank holder ring 24 structure during the secondary drawing process, providing a full-range, balanced, adaptive flexible blank holder force for the component blank after the primary drawing. Its working process is as follows: when the die 22 moves downwards, the nitrogen spring 7 pushes the elastic pressure plate to contact the component first and provides a progressive blank holder force as the stroke increases, effectively constraining material flow and preventing wrinkling; when the die 22 returns, the nitrogen spring 7 drives the pressure plate to quickly reset. This simplifies the die structure and is particularly suitable for secondary finishing drawing processes requiring minor supplementary forming.

[0068] Optionally, in the above technical solution, the sheet metal forming defect risk perception and edge clamping force self-feedback control system also includes a light-emitting device, which is arranged side by side with the image acquisition component 44 on the gimbal 6.

[0069] In this technical solution, by setting up a light-emitting device, the brightness of the plate 23 can be increased, thereby improving the image acquisition effect.

[0070] like Figure 5As shown, the second aspect of the present invention provides a control method for a sheet metal forming defect risk perception and blank holder force self-feedback control system, comprising:

[0071] S202: After the current forming cycle ends, the outer contour feature image of the sheet metal is acquired through the image acquisition component;

[0072] S204: Analyze the defect risk type in the outer contour feature image of each detection area using an image analysis device, and generate a pressure compensation command based on the defect risk type;

[0073] S206: During the time window from the end of the current forming cycle to the start of the next cycle, a pressure compensation command is sent to the hydraulic device in the corresponding detection area via the industrial bus to adjust the blank holder force for the next forming cycle.

[0074] The control method provided by this invention includes: in each forming cycle, when the die 22 returns to the top dead center (i.e., moves to the topmost position until it can no longer move), the defect risk type in the outer contour feature image of each detection area is analyzed by an image analysis device, and a pressure compensation command is generated according to the defect risk type. Specifically, for the wrinkling defect risk, a pressure increase command (amplitude +5%~+15% of the current blank holder force) is generated to suppress material flow and accumulation; for the cracking defect risk, a pressure reduction command (amplitude -5%~-15% of the current blank holder force) is generated to reduce the peak tensile stress. The pressure increase / depression adopts a step-by-step adjustment method, with each step adjustment amplitude of 3%–5%. After each adjustment, the change value of the take-up line size within 10-30 strokes is monitored to determine whether it has returned to the safe range. If it has not returned, the next incremental step adjustment needs to be continued until the take-up line size is restored to the safe range. Within the time window from the end of the current forming cycle to the start of the next cycle, the pressure compensation command is sent to the corresponding hydraulic device 42 through the industrial bus to achieve precise control of the blank holder force in the next forming cycle. This method achieves zoned pressure compensation, overcomes the problems of wrinkling and cracking, and completely breaks through the technical bottleneck of difficulty in defect control and unstable forming quality caused by the lag in response of traditional edge pressure adjustment methods.

[0075] For example, as follows Figure 6As shown, through simulation and combined with on-site busbar debugging, the simulation results are corrected and optimized to determine that points A and B of sheet 23 are prone to defects. Suppose that point A falls within the first detection area 32 and point B falls within the second detection area 34. The pressing ring 24 has pressing sections in the first detection area 32 and the second detection area 34 respectively, and each pressing section is equipped with a hydraulic device 42. After the previous sheet 23 is drawn and formed, the image acquisition component 44 acquires the outer contour feature images of the sheet 23 in the two detection areas and analyzes them through the image analysis device. The main analysis is the distance between points A and B before and after forming, so as to determine whether there is a risk of cracking or wrinkling. If there is no defect risk in the sheet 23 of both detection areas, the device will not operate. If there is a defect risk in the sheet 23 of the first detection area 32, the pressure of the pressure ring 24 of the first detection area 32 will be adjusted accordingly to ensure the normal drawing and forming of the next sheet 23. Similarly, if there is a defect risk in the sheet 23 of the second detection area 34, the pressure of the pressure ring 24 of the second detection area 34 will be adjusted. A risk detection will be performed after each forming cycle to ensure that all sheets 23 meet the standards.

[0076] It is important to understand that there may be one or multiple defect risk points within the same inspection area, and the defect risk points are determined based on the component.

[0077] In the above technical solution, the image acquisition component includes a camera and a lens. The steps of analyzing the defect risk type in the outer contour feature image of each detection area using an image analysis device include: constructing a finite element simulation model based on the property parameters of the sheet metal; performing multi-condition simulation based on the finite element simulation model, determining the first detection point on the sheet metal, and obtaining the take-up line length L of the first detection point under ideal forming conditions. 理 The length L of the take-up line at the critical state of wrinkling. 皱 And the length L of the take-up line at the critical cracking state. 裂 The simulation results were corrected and optimized based on the on-site busbar commissioning, and a safe critical size range for the take-up line was finally set for the first detection point [L]. min L max The image acquisition component is calibrated using a checkerboard calibration board, generating a transformation matrix from pixel coordinates to physical space coordinates. The position corresponding to the first detection point is determined in the outer contour feature image and denoted as the second detection point. The pixel distance from the second detection point to the outer edge of the sheet is calculated. This pixel distance is converted to physical space coordinates using the transformation matrix, and the actual take-up line length L of the second detection point is calculated. 实际 ; will L 实际 The critical size range of the receiving line [L] min L maxCompare: If L 实际 < L min If L 实际 > L max If the test result is positive, then the area where the test point is located is deemed to be at risk of cracking.

[0078] In this technical solution, the step of analyzing defect risk types using an image analysis device in the control method includes: constructing a finite element simulation model based on the property parameters of sheet metal 23; then, performing multi-condition simulation based on the finite element simulation model to determine the first detection point on the sheet metal and obtain the take-up line length L of the first detection point under ideal forming conditions. 理 The length L of the take-up line at the critical state of wrinkling. 皱 And the length L of the take-up line at the critical cracking state. 裂 The first detection point, which is also the detection point where risks are likely to occur in the finite element simulation model (i.e., the aforementioned defect risk-prone point), is the detection point for risk comparison in the later stage. The receiving line, which is the distance before and after forming of the detection point, is determined by multi-condition simulation experiments. 皱 and L 裂 Subsequently, the simulation results were corrected and optimized based on the on-site busbar commissioning, and a safe critical size range for the take-up line was finally set for the first detection point [L]. min L max ], where L min >L 皱 L max < L 裂 When the receiving line is greater than L 裂 When this happens, it indicates that the sheet metal has already cracked, therefore a safety margin L needs to be set. max Similarly, when the receiving line is less than L 皱 When this occurs, it indicates that the sheet metal has already wrinkled, therefore a safety margin L needs to be set. min That is, the receiving line is at L min and L max If the values ​​are within the acceptable range, the molded part is considered acceptable; if they are outside the acceptable range, it indicates a risk of cracking or wrinkling.

[0079] In determining L min and L maxDuring the process, the primary and secondary strains of key feature points are first extracted and compared with the forming limit diagram (determined by Nakazima experiments). If the strain state of a feature point is within the 90% to 95% limit strain range of the forming limit curve, it is considered that the point has reached the critical forming state. Based on this, the corresponding take-up line size is determined as the critical size, and the simulation results are corrected and optimized in conjunction with on-site busbar debugging to determine the safe size range of the entire take-up line. Furthermore, after actual forming and acquisition of the outer contour feature image of the sheet metal, the actual take-up line is calculated. The actual take-up line is determined as follows: the image acquisition component is calibrated using a checkerboard calibration plate to generate a transformation matrix from pixel coordinates to physical space coordinates; the same position as the first detection point is determined in the outer contour feature image and recorded as the second detection point; that is, the second detection point and the first detection point are the same point on the sheet metal 23, thereby ensuring the consistency between the detection points of the finite element simulation model and the actual detection points, and thus making a risk type judgment.

[0080] Specifically, locating the "second detection point" on the outer contour feature image can be achieved through a combination of feature matching and coordinate mapping. First, in the finite element simulation model, the position of the first detection point is defined based on the relative position of fixed geometric features on the mold (such as the center of the punch fillet, the intersection of specific edges of the die, or a preset laser marking point), and its coordinates are determined in the world coordinate system of the simulation software.

[0081] Before actual forming, a checkerboard calibration plate is fixed to the mold surface or a platform with a known relative position to the mold to calibrate the image acquisition component. This step simultaneously accomplishes two tasks: generating a transformation matrix from pixel coordinates to physical space coordinates (for receiving line dimension measurement) and establishing a mapping relationship between the image pixel coordinate system and the simulation software world coordinate system. After actually acquiring the sheet metal image, image processing algorithms are used to automatically identify the fixed geometric features of the mold in the image, which serve as a reference. Based on the established mapping relationship between the image pixel coordinate system and the simulation software world coordinate system, the coordinates of the first detection point in the simulation world are transformed to the pixel coordinate system of the current image, and the theoretical pixel coordinates of the second detection point are calculated.

[0082] After determining the second detection point, calculate the pixel distance from the second detection point to the outer edge of the sheet material; use a transformation matrix to convert it to physical space coordinates, and calculate the actual take-up line distance L of the second detection point. 实际 ; will L 实际 The critical size range of the receiving line [L] min L max Compare: If L 实际 < L min If L实际 > L max If the test result is positive, then the area where the test point is located is deemed to be at risk of cracking.

[0083] Furthermore, there can be multiple detection points, that is, there are multiple first detection points and multiple second detection points, and the number of first detection points and second detection points is the same. The actual material receiving line for each detection point is determined in the above manner. There can be one detection point or multiple detection points on the same detection area. When there are multiple detection points on the same detection area, if one detection point fails, the blank holder force adjustment command for the blank holder section corresponding to that detection area is triggered to ensure that the material flow in that area returns to normal in the next forming cycle, and ultimately ensure the overall forming quality of the component.

[0084] Among them, after multiple strokes, when the component is at risk of cracking, the changes in the take-up line are as follows: Figure 10 As shown, when there is a risk of wrinkling in the component, the changes in the take-up line are as follows: Figure 11 As shown, when the component is within the safe range, the changes in the receiving line are as follows: Figure 12 As shown. Figures 10 to 12 In this context, the safety limit is also known as L in this application. max The safety lower limit is also known as L in this application. min .

[0085] Furthermore, the receiving line control principle of this application is as follows: Figure 6 As shown, if during the experiment using the finite element simulation model, adjusting the blank holder force proves that points A and B are prone to cracking or wrinkling, then the device will automatically identify points A and B as the first detection points. That is, the first detection points are determined based on the finite element simulation and, combined with on-site busbar debugging, by correcting and optimizing the simulation results. Further, the critical size range of the receiving line [L] is determined. min L max In this context, the take-up line refers to the distance between the detection point and the forming point. For example, taking detection point B as an example, in the simulation experiment, point B is in an ideal position, representing the most standard component. Then, by increasing the blank holder force, the component is made to crack. If the component cracks when detection point B moves to point B1, the distance from point B1 to the outer edge of the component is recorded as the limit dimension of the take-up line. This means that in the actual production process, if the take-up line is greater than this limit dimension, it indicates that the component has cracked. Therefore, in the actual production process, it is only necessary to ensure that the actual size of the take-up line is within [L]. min L max The interval is acceptable, when L 实际 >L max If the risk of cracking is identified, the blank holder force should be reduced until L... 实际 <Lmin If the condition is deemed to have a risk of wrinkling, the edge pressure should be increased.

[0086] In the above technical solution, optionally, the step of constructing a finite element simulation model based on the property parameters of sheet metal 23 specifically includes: constructing a three-dimensional geometric model in simulation software that includes sheet metal 23, die 22, punch 21 and blank holder 24 according to the geometric shape and structural dimensions of sheet metal 23; setting a material constitutive model for sheet metal 23; defining the contact relationship between sheet metal 23 and die 22 and punch 21, and setting the friction coefficient; applying the blank holder force parameter as a boundary condition to the contact surface between blank holder 24 and sheet metal 23; and integrating the three-dimensional geometric model, material constitutive model, contact relationship and boundary condition through a finite element solver to generate a finite element simulation model for simulating forming processes with different blank holder forces and friction coefficients.

[0087] Another embodiment of the present invention provides a control method for a sheet metal forming defect risk perception and blank holder force self-feedback control system. This method mainly enables risk perception of sheet metal forming defects and self-feedback control of the blank holder force. It is applicable to defect early warning and quality control in the primary and secondary drawing processes of metal sheets, and mainly includes the following steps:

[0088] S1. Establish a finite element simulation model for the target sheet 23. Input process parameters including the stress-strain curve of sheet 23, anisotropic yield criterion, forming limit curve (determined by Nakazima test), friction coefficient and mold profile, etc. Execute multi-condition simulation to obtain the influence of different blank holder forces on the evolution process of the take-up line. Combine the on-site busbar debugging to correct and optimize the simulation results. Finally, output the critical dimensions of the take-up line and its allowable fluctuation threshold range of the key defect-prone area at different forming stages.

[0089] S2. In the actual production process, high-precision image acquisition of the formed sheet 23 is achieved through synchronous camera, lens and anti-interference device (i.e. the above-mentioned image acquisition component 44), so as to ensure stable image quality and minimal interference of the receiving line.

[0090] S3. Using the image processing and size calculation module (i.e. the image analysis device mentioned above), multi-scale edge fusion processing is performed on the acquired image. Combined with contour detection algorithm / deep learning, the material receiving line boundary is accurately extracted. And the image data is converted into actual size information with millimeter-level accuracy through calibration mapping.

[0091] S4. Using the risk identification module, the system collects the dimensional data of key points on the receiving line in real time, performs online fitting on the data points, and then compares and analyzes them with the threshold range of the receiving line to automatically identify the type of defect risk and trigger an early warning signal. At the same time, it provides fault-tolerant logic to prevent misjudgment.

[0092] S5. Based on the risk identification results, the force values ​​of the hydraulic cylinders / nitrogen springs (i.e., the hydraulic devices 42 mentioned above) in the corresponding areas of the primary and secondary drawing processes are corrected to achieve local closed-loop control only for the risk areas, thereby improving the adjustment accuracy and system response efficiency.

[0093] S6. Utilizing the central controller and bus coordination module, it enables high-frequency and high-reliability data interaction between various structures, unifies task scheduling, synchronizes status, and ensures stable system operation, while also possessing compatibility with multi-protocol industrial networks.

[0094] S7. Utilizing human-machine interaction and data archiving devices, the system allows for real-time monitoring of risk distribution, dimensional changes, and adjustment status. Simultaneously, the system automatically archives and traces process data such as inspection images, edge clamping force responses, and early warning logs, generating batch quality reports to support quality tracking and process optimization.

[0095] Furthermore, step S1 specifically includes:

[0096] (1) Establish simulation geometric model and material model

[0097] a) Based on the geometry and structural dimensions of the target sheet 23, construct a three-dimensional geometric model in the simulation software, including the sheet 23, the die 22, the punch 21, and the pressure ring 24, etc.

[0098] b) Set a reasonable material constitutive model for sheet 23, such as elastic model, hardening model, anisotropic yield model, forming limit curve (determined by Nakazima test), etc., and fully consider the yield behavior and plastic flow characteristics of the material.

[0099] (2) Setting process boundary conditions and contact behavior

[0100] a) Define the contact relationship and friction coefficient, especially the contact area between sheet 23 and the mold, to ensure the realism of frictional slip and force transmission in the simulation;

[0101] b) Set the loading path, including the stroke of slider 5, the lifting height of pressure ring 24, the application method of pressure force (fixed or regional variable pressure), drawing speed, etc.

[0102] c) Import the mold geometry and define its kinematic properties to ensure that the relative motion between the mold and the sheet 23 conforms to the actual working conditions.

[0103] (3) Generating finite element mesh

[0104] a) Based on the geometric model and the set boundary conditions, mesh the sheet metal 23 and the mold, and select appropriate element types and sizes;

[0105] b) For easily deformable areas (such as rounded corners and near the receiving line), a finer mesh is used to improve simulation accuracy; for other areas, a relatively coarse mesh is used to optimize computational efficiency.

[0106] c) Ensure mesh continuity and accuracy in the contact area and critical loading path to avoid numerical noise.

[0107] (4) Simulation

[0108] a) Set up multiple working condition combinations, including different blank holder forces, friction coefficients, etc., to simulate the forming process under different working conditions;

[0109] b) Define appropriate boundary conditions for the simulation model to ensure that the loading process is consistent with the actual working conditions;

[0110] c) Start simulation calculations to simulate the evolution of the forming process under different working conditions, focusing on the changes in the receiving line and the deformation process.

[0111] (5) Analyze the effect of blank holder force on the take-up line.

[0112] Observe the changes in the take-up line under different blank holder forces. Excessive or insufficient blank holder force may lead to defects in the sheet material, such as cracking and wrinkling.

[0113] (6) Output the critical defect-prone areas and the critical dimensions of the receiving line.

[0114] a) Based on simulation results, risk areas that may lead to defects due to uneven deformation at different forming stages are identified;

[0115] b) Extract the principal and secondary strains of key feature points and compare them with the forming limit curve (determined by Nakazima experiments). If the strain state of a feature point is within the 90%–95% limit strain range of the forming limit curve, then the point is considered to have reached the critical forming state. Based on this, determine the corresponding take-up line size as the critical size, and combine it with on-site busbar debugging to correct and optimize the simulation results, thereby determining the safe size range of the entire take-up line.

[0116] Furthermore, step S2 specifically includes:

[0117] (1) System hardware deployment

[0118] a) Install n sets (based on the number of hydraulic cylinders / nitrogen springs) of high-resolution cameras above the receiving line of the forming equipment to achieve multi-angle, full-coverage observation, for example, five sets.

[0119] b) To improve image clarity and boundary contrast, coaxial or backlight sources are installed, and the light source is synchronized with the camera to avoid interference from shadows and reflections.

[0120] c) Based on the shape of the mold and the geometric characteristics of the sheet metal 23, adjust the installation angle and focal length of each camera group to ensure that the imaging area completely covers the receiving line and avoids blind spots.

[0121] (2) Camera dynamic attitude control and adaptation mechanism

[0122] a) Each camera group is equipped with an electronically controlled sliding rail structure, which supports linear movement of the camera in the horizontal and vertical directions. Multi-dimensional position adjustment is achieved through electric drive components to control the position of the camera body and adapt to the layout requirements of different monitoring areas.

[0123] b) To cope with different mold structures and changes in the geometric distribution of the receiving line, each camera is mounted on a precision gimbal 6 with dual-axis or three-axis servo control capabilities, enabling electronically controlled adjustments for angle tilt and position fine-tuning, thereby improving the shooting accuracy and angle coverage of the target area.

[0124] c) The camera monitoring parameters are linked with the mold parameter database. Based on the current mold number or part specification, the corresponding field of view configuration file is automatically loaded, and the servo control logic of the gimbal 6 is executed to quickly complete the pre-adjustment and positioning of the camera attitude.

[0125] d) To ensure that the adjusted imaging area accurately covers the effective monitoring area, the system corrects the position error of the camera's current viewing angle by capturing a standard calibration pattern or a fixed reference object.

[0126] e) The camera attitude adjustment process is managed by the central controller and automatically coordinated with the image acquisition task to ensure that the image acquisition trigger time delay does not exceed 50ms after the adjustment is completed, so as to avoid production cycle interruption caused by mold change.

[0127] f) The servo gimbal 6 system supports remote maintenance and manual intervention. The attitude parameters (pitch angle, rotation angle, translation displacement, etc.) of each camera can be viewed in real time through the human-machine interface, and a one-click reset and rollback strategy is provided to improve system maintenance efficiency.

[0128] (3) Establish a timing triggering and synchronization mechanism

[0129] a) Establish signal communication with the main control system of the forming equipment and set the triggering time for image acquisition (such as judging based on the position of the die 22 running to the top dead center, or judging based on the position signal output by the encoder).

[0130] b) All cameras are managed by a central controller, which enables multi-channel synchronous image acquisition based on timestamps or external hardware trigger signals.

[0131] (4) Ensure image stability and environmental interference resistance

[0132] a) All cameras and light source devices are equipped with industrial-grade dustproof, vibration-proof, and electromagnetic interference-proof housings to adapt to the field working environment.

[0133] b) Configure an intelligent temperature control module inside the camera housing to keep the image acquisition device operating within a stable temperature range and prevent overheating from causing imaging abnormalities.

[0134] c) The optical self-calibration function of the image device is triggered periodically or periodically to compensate for image deviations caused by light, temperature, and position fine-tuning, thereby improving long-term acquisition stability.

[0135] (5) Image acquisition and transmission process

[0136] a) Acquire images in a high-resolution, lossless format to preserve image details, especially in areas used for subsequent edge extraction processing.

[0137] b) Transmit image data to the image processing and size calculation module in real time via gigabit Ethernet or fiber optic interface to ensure uncompressed and delay-free image transmission.

[0138] c) Set up local caching and data integrity verification mechanisms to avoid image data loss or incompleteness due to network fluctuations or sudden anomalies.

[0139] (6) Image acquisition quality monitoring mechanism

[0140] a) The system evaluates the contrast, sharpness, and illumination uniformity of the acquired images. If the image quality is lower than the set threshold, the system will automatically trigger image re-acquisition or issue a quality warning.

[0141] b) Continuously monitor the working status of each camera (such as exposure time, image clarity, temperature, voltage, etc.), and automatically record logs and push them to the human-machine interface in case of any abnormality.

[0142] (7) Data collection archiving and identification

[0143] a) Each frame of image is automatically bound with information such as sheet metal 23 identifier, punch number, timestamp, and camera number to ensure subsequent processing and quality tracking.

[0144] b) The acquired images and metadata are simultaneously transmitted to the data archiving module for categorized storage, and can be quickly retrieved by batch, working condition, time and other conditions.

[0145] Furthermore, step S3 specifically includes:

[0146] (1) Image preprocessing and standardization

[0147] a) Convert the images captured by the camera from their original format to a standard processing format, and perform grayscale or brightness equalization processing to eliminate image interference caused by changes in lighting.

[0148] b) Apply methods such as average filtering and Gaussian filtering to reduce noise in the image, remove unstructured noise such as dust and scratches, and ensure the stability of subsequent edge detection.

[0149] c) Based on the relative positioning of sheet metal 23 in the production cycle, the effective area containing the receiving area in the image is automatically cut, shortening the processing time and improving the analysis efficiency.

[0150] (2) Multi-scale edge blending processing

[0151] a) An edge detection algorithm is used to extract the contour of the potential receiving line at different image scales.

[0152] b) Introduce image pyramid structure processing technology or use multi-resolution image fusion methods to construct a full-scale edge map containing coarse and detailed boundaries by integrating edge response information at different scales.

[0153] c) By normalizing the edge response amplitude and suppressing non-maximum values, the target boundary is strengthened while stray edges are suppressed.

[0154] (3) Extraction and vectorization of receiving line boundary

[0155] a) Use contour detection algorithms to extract continuous boundaries and fit the boundary points (e.g., least squares line fitting, circular arc fitting, spline interpolation, etc.) to form smooth geometric curves. Alternatively, select a lightweight boundary detection model suitable for industrial inspection (such as HED, DeepEdge, etc.) and perform transfer learning and fine-tuning based on actual image features.

[0156] b) Using existing sheet metal image samples, construct a training dataset by manually annotating the true boundaries of the receiving line, and enhance sample diversity (rotation, noise reduction, occlusion).

[0157] c) Input the edge enhancement image into the deep learning model to obtain a high-confidence pixel-level mask image of the receiving line, and automatically filter out false boundaries and artifacts.

[0158] d) Automatically extract inflection points, peaks and troughs, and maximum offset points in the boundary to facilitate subsequent dimensioning.

[0159] (4) Mapping transformation from image coordinates to physical dimensions

[0160] a) Use a checkerboard or dot array calibration board, combined with standard camera calibration methods, to obtain camera intrinsic parameters (focal length, principal point, distortion coefficient, etc.) and extrinsic parameters (attitude matrix, etc.).

[0161] b) Construct a homography transformation matrix from image pixels to physical space (millimeter level) coordinates to achieve unbiased restoration of pixel boundaries to actual boundaries.

[0162] c) Perform radial and tangential distortion correction on the image to ensure the linearity and proportionality accuracy after boundary coordinate transformation, thereby improving the accuracy of physical measurements.

[0163] d) Calculate the actual spacing between continuous boundary points to obtain key dimensional indicators such as the actual length, curvature, and fluctuation amplitude of the receiving line.

[0164] (5) Generation of Dimension Output and Visualization Results

[0165] a) Overlay the receiving line boundary vector and the corresponding physical dimension value onto the image, and highlight abnormal areas with a red warning box.

[0166] b) Output the key point dimension data of the receiving line (maximum deviation, fluctuation range, extreme point) in CSV or JSON format.

[0167] c) Synchronize the processing results to the human-computer interaction interface to provide intuitive and readable information such as the material receiving line size trend chart, stability assessment indicators, and historical fluctuation trajectory.

[0168] Furthermore, step S4 specifically includes:

[0169] (1) Real-time data acquisition

[0170] Using image processing and size calculation modules, the system continuously acquires size data (such as edge width, centering error, etc.) of key parts of the receiving line, and the data sampling cycle is synchronized with the production rhythm.

[0171] (2) Input the threshold range of the receiving line

[0172] a) The system pre-introduces the threshold range for the receiving line.

[0173] b) The threshold range takes into account the reasonable dimensional fluctuation range under different operating conditions (such as speed, load, etc.).

[0174] c) Supports dynamic adjustment of upper and lower thresholds based on time series and operating condition switching.

[0175] (3) Comparison and difference analysis

[0176] a) Compare the real-time collected key point size data with the threshold range under the current state one by one.

[0177] b) Calculate the deviation and trend between the actual size and the threshold range.

[0178] c) Analyze the continuous trend of the deviation amplitude to improve the stability and anti-interference of the judgment.

[0179] (4) Risk type judgment

[0180] Using the risk identification module, the dimensional data of key points on the receiving line are collected in real time, and the data points are fitted online. Then, the data is compared and analyzed with the threshold range of the receiving line to automatically identify the type of defect risk.

[0181] (5) Warning signal triggered

[0182] Based on the identified risk type, the system issues corresponding early warning signals through multiple channels (such as audible and visual alarms, system pop-ups, and background push notifications), and pushes risk information to operator terminals or operation and maintenance platforms to achieve rapid linkage response.

[0183] (6) Fault tolerance and misjudgment suppression mechanism

[0184] a) Set judgment conditions such as minimum duration or number of consecutive abnormal frames to avoid false alarms caused by instantaneous errors.

[0185] b) Introduce statistical filtering or logical verification mechanisms to conduct secondary confirmation of suspected anomalies.

[0186] Furthermore, the primary and secondary drawing sequences are each equipped with independent blank holder force self-feedback adjustment modules.

[0187] In one drawing sequence, a servo-controlled hydraulic cylinder is used as the blank-pressing execution unit. A closed-loop regulation circuit is constructed by integrating an electro-hydraulic servo valve and a pressure sensor in the hydraulic circuit of each hydraulic cylinder. The electro-hydraulic servo valve is used to dynamically adjust the oil circuit pressure and flow according to the target blank-pressing force adjustment command output by the control system. The pressure sensor collects the actual oil pressure in the cylinder in real time, and compares the error with the target value through the controller to drive the servo valve to make continuous corrections, thereby realizing high-precision dynamic closed-loop control of the blank-pressing force.

[0188] In the secondary drawing sequence, the blank holder execution unit adopts an interface-type nitrogen spring cylinder group. Through a centralized air supply structure with a standard air source interface, it achieves air pressure regulation and centralized control. Proportional electronically controlled pressure regulating valves and pressure feedback sensors are installed in the air supply pipelines of each cylinder group to construct a closed-loop regulation and control circuit. The proportional electronically controlled pressure regulating valves quickly adjust the air supply pressure of the target area cylinder group according to the blank holder force adjustment command output by the control system. The pressure sensor collects the current air pressure data of the cylinder group in real time and compares it with the target value. The control system adaptively adjusts according to the pressure error, driving the proportional valve to output a correction signal, thus achieving rapid closed-loop regulation of the blank holder force. By optimizing the air path diameter, improving the air supply response capability and valve control accuracy, it ensures that the blank holder force completes dynamic variable adjustment within a single stroke cycle, thereby meeting the requirements for rapid response and adaptive adjustment of the blank holder force in different risk areas under high-speed cyclic conditions.

[0189] Furthermore, step S5 specifically includes:

[0190] (1) Risk area input identification

[0191] Receive the risk type determination results for each monitored area from the risk identification module.

[0192] (2) Preparation of blank holder force adjustment parameters

[0193] Based on the risk type, set the corresponding target blank holder force adjustment amount or adjustment ratio. For the risk of wrinkling defects, generate an increase command (amplitude +5% to +15% of the current blank holder force) to suppress material flow and accumulation; for the risk of cracking defects, generate a decrease command (amplitude -5% to -15% of the current blank holder force) to reduce the peak tensile stress. The increase / decrease is adjusted in steps, with each step adjusting by 3%–5%. After each adjustment, monitor the change in the take-up line size within 10-30 strokes to determine if it has returned to the safe range. If it has not returned, continue to the next incremental step adjustment until the take-up line size returns to the safe range. At the same time, the recommended blank holder force range in the simulation model or expert rule base is considered as the limiting boundary.

[0194] (3) Hydraulic cylinder / nitrogen spring selection area and closed-loop control activation

[0195] a) Activate the corresponding hydraulic cylinder / nitrogen spring control circuit according to the coordinate mapping of the risk identification area.

[0196] b) Maintain the original control status for non-target areas (those that have not triggered risks) to avoid increasing system load or misadjustment caused by full-area intervention.

[0197] (4) Real-time feedback collection and adjustment execution

[0198] a) Initiate local closed-loop control to adjust the force value in the target area.

[0199] b) The current actual blank holder force is fed back by the pressure sensor and compared with the adjustment target value to perform PID adaptive control adjustment.

[0200] c) Based on the information received from the risk identification module, generate a dimension diagram of key points of the stroke and take-up line, and perform online fitting of the data to further determine the effect of adjusting the blank holder force on improving the dimensions of the take-up line.

[0201] (5) Fault-tolerant and security logic design

[0202] a) Set adjustment upper / lower limit protection values ​​to prevent misidentification leading to overvoltage or undervoltage.

[0203] b) Supports manual intervention or adjustment of authorization mechanisms, and can quickly switch to manual confirmation mode when the system misjudges.

[0204] c) All adjustment records and risk identification data are automatically archived for use in quality analysis and retrospective analysis.

[0205] Furthermore, the central controller and bus coordination module are used to: realize real-time data communication between the vision inspection system, the blank holder force adjustment system, and the simulation model database; provide system scheduling, task synchronization, and redundant fault-tolerant control logic; and support high-frequency communication protocols based on industrial fieldbus or real-time Ethernet, including but not limited to EtherCAT, PROFINET, etc., to realize real-time data synchronization and control command transmission between various modules of the system.

[0206] Furthermore, step S6 specifically includes:

[0207] (1) Main controller initialization

[0208] a) Upon startup, the central controller loads the communication configurations of all modules (module ID, transmission cycle, priority, etc.).

[0209] b) Establish a main task scheduling table and define periodic tasks, event-driven tasks, and background management tasks.

[0210] (2) Bus network identification and configuration

[0211] a) Automatically detect the type of the currently connected industrial bus.

[0212] b) Call the multi-protocol adaptation layer, load the corresponding communication protocol stack, and complete the address binding and protocol mapping with each module.

[0213] c) Establish a bus mapping table for each module, recording its physical port, logical identifier, and status register address.

[0214] (3) High-frequency data interaction scheduling

[0215] a) Set the communication cycle for each module task according to priority and sampling frequency.

[0216] b) Start the timed interrupt polling mechanism and initiate data collection and command issuance according to the schedule.

[0217] (4) Unified state synchronization and task scheduling

[0218] a) Collect the operating status and data packets of all sub-modules (sensors, actuators, analysis modules, etc.).

[0219] b) The main controller dynamically allocates resources based on the current operating conditions and module status, such as the update frequency of hydraulic cylinder / nitrogen spring force values ​​and the priority of early warning responses.

[0220] c) State changes are broadcast to all modules via a global state machine to achieve logical consistency and task coordination.

[0221] (5) Fault detection and communication monitoring

[0222] a) Real-time monitoring of communication anomalies (such as frame loss, CRC error, response timeout, etc.).

[0223] b) When an anomaly occurs, mark the module status as "offline / disconnected" to trigger a backup strategy or redundancy mechanism.

[0224] c) All communication events and system log records are automatically archived, and remote diagnostic reporting is supported.

[0225] (6) System fault tolerance and hot-plug support

[0226] a) Supports hot-swappable modules or online start / stop: The controller automatically detects module connection / disconnection and dynamically reconfigures the communication link.

[0227] b) After the module restarts, it can achieve fast and seamless recovery through state synchronization and task recovery mechanisms.

[0228] (7) Multi-protocol compatibility mechanism

[0229] a) Introduce a protocol gateway module or a unified communication interface.

[0230] b) The system is compatible with modules from different suppliers and protocols of various industrial equipment, enabling heterogeneous network integration.

[0231] c) Enables protocol switching and parameter adjustment through configuration interfaces or scripts, improving system scalability and deployment flexibility.

[0232] (8) Security and bandwidth management strategy

[0233] a) Enable bus bandwidth allocation control to prevent high-frequency communication of a single module from consuming all bus resources.

[0234] b) Encrypt critical data segments (such as control commands and risk signals) to prevent malicious external interference.

[0235] c) Implement identity authentication and communication permission table mechanisms to ensure that the system control security boundaries are clear and controllable.

[0236] Furthermore, step S7 specifically includes:

[0237] (1) Data acquisition and real-time monitoring

[0238] a) The risk distribution of each monitoring area is obtained in real time through the data acquisition module, and the risk identification results are displayed through a graphical interface, which supports dynamic updates.

[0239] b) Collect dimensional change data during the forming process of the formed parts through a multi-channel industrial vision inspection module, monitor the change trend of key dimensions in real time, and display information such as dimensional deviation and change rate, and present it through trend lines or real-time data tables, so that the operator can track it in real time.

[0240] c) The system displays the real-time status of the blank holder force adjustment, showing the actual blank holder force value, target value, and deviation for each area. If the adjustment deviation exceeds the limit or the status is abnormal, the system will automatically highlight the error message to facilitate timely adjustments by staff.

[0241] (2) Data archiving and process recording

[0242] a) All process data, including detection images, edge clamping force response, early warning logs, and adjustment status, are automatically archived by timestamp, batch number, etc. File categorization storage (e.g., log files, sensor data files, image files) is supported, as well as filtering and retrieval by date, batch, and module.

[0243] b) When an abnormal warning occurs (such as exceeding the limit of the risk area, sensor failure, etc.), the system automatically records the event details (time, module, abnormality type, response measures, etc.) and generates a warning log. The log file is uploaded to the database in real time to ensure the accuracy of event tracing.

[0244] (3) Quality report generation and batch tracking

[0245] a) After each production or testing cycle, the system automatically generates a complete process data report for that batch, including all relevant data, status changes, risk assessment results, etc. The data is linked according to batch number, production date and other information to ensure data integrity and traceability.

[0246] (b) The system generates batch quality reports based on archived data and real-time status, including: component size and tolerance assessment; control effectiveness assessment for each risk area; blank holder force adjustment status and response analysis; and abnormal events and response measures during the production process. After generating the report, it can be exported as PDF or Excel format and sent directly to quality management personnel, production managers, and other relevant personnel through the interface.

[0247] (4) Data visualization and report presentation

[0248] a) The human-machine interface provides a large visual display area, which displays key data such as real-time risk distribution, size changes, and edge clamping force response through charts, data curves, etc.

[0249] b) Users can use the query function to select specific batch numbers, date ranges, module types, etc., to query historical data.

[0250] c) Supports multi-dimensional data filtering to quickly locate specific quality problems or adjust anomalies.

[0251] (5) System alarms and abnormal responses

[0252] a) During the data acquisition process, the system monitors in real time according to the set risk threshold. Once an anomaly is detected (such as excessive or insufficient blank holder force, or dimensions exceeding tolerance), the system immediately issues an alarm, automatically records the alarm content, and triggers the corresponding early warning log and data archiving.

[0253] (b) After each alarm, the system automatically records the response measures and adjustment plans, and archives them as part of the historical data. Through the alarm module, users can view detailed alarm history to understand the system's operating status and processing procedures.

[0254] (6) Process optimization support

[0255] a) Conduct trend analysis using historical process data and quality reports to assess the impact of different adjustment strategies and risk control measures on quality, and regularly generate process optimization recommendation reports.

[0256] b) Based on historical data analysis results, the system can automatically adjust parameters and optimize the production process according to the production situation, and provide real-time feedback to the operators. Based on the system feedback, the system can also adjust the process plan and parameter configuration to further improve product quality.

[0257] Another embodiment of the present invention provides a control method for a sheet metal forming defect risk perception and blank holder force self-feedback control system. This method primarily enables risk perception of sheet metal forming defects and self-feedback control of the blank holder force. It deeply integrates finite element simulation analysis, multi-channel industrial vision inspection, image boundary recognition algorithms, and closed-loop adjustment of the blank holder force region, constructing a method and device with real-time defect early warning and adaptive blank holder force adjustment capabilities. Based on multi-condition simulation results and combined with on-site busbar debugging, the system corrects and optimizes the simulation results, generating critical dimensions and fluctuation thresholds for the take-up line in key areas. Combined with a high-precision image acquisition component 44 and boundary recognition algorithms, it dynamically monitors the changing trend of the take-up line, assesses the risk of wrinkling, cracking, and other defects in the formed parts in real time, and implements rapid closed-loop correction of the blank holder force in the corresponding areas. This upgrades the system from post-detection to an intelligent closed-loop control mode integrating defect early warning and real-time control. Compared to traditional quality inspection methods that rely on manual sampling and quality control methods that adjust process parameters based on experience, this invention can simultaneously achieve accurate identification, dynamic early warning, and real-time control of sheet metal forming defects on-site. It has a higher ability to ensure forming consistency and adapt to production cycle time. It not only improves the stability of the forming process and product quality, but also significantly reduces the degree of manual intervention and debugging costs. It has excellent industrial site adaptability, system compatibility, and prospects for widespread application.

[0258] The first step is to use the finite element simulation module to establish a forming finite element model for the target sheet 23. The input includes material stress-strain relationship, anisotropic yield criterion, forming limit curve (determined by Nakazima test), friction coefficient, blank holder force and other material properties and boundary conditions. Multi-condition simulation is performed to obtain the influence of different blank holder forces on the evolution process of the take-up line. The simulation results are corrected and optimized in combination with on-site busbar debugging. The critical dimensions of the take-up line and its allowable fluctuation threshold range of the key defect-prone area at different forming stages are output.

[0259] Specifically, taking the side panel of a car body panel as an example, the 3D model of the side panel and its matching forming mold is completed using Catia software. After modeling, the geometry is preprocessed, including simplifying chamfer features, merging small faces, and repairing gaps, to ensure the continuity of the model and its compatibility with finite element analysis. After processing, the CAD model is exported in STEP format and imported into Abaqus software for finite element preprocessing analysis. In Abaqus, the imported geometric model is identified as a part, with the side panel of the car body panel set as a deformable body and the forming mold set as a rigid body. Body); Define the physical parameters of the side panel material of the body panel, where the elastic modulus is set to 70000. MPa, Poisson's ratio 0.33, develop VUMAT subroutine, adopt Hockett-Sherby hardening model (e.g. Figure 7 As shown), BBC2005 yield model (as shown) Figure 8 As shown, σ (for stress) and material forming limit curve (such as) Figure 9 (As shown in the figure). In terms of process settings, forming process boundary conditions are set, including motion trajectory control of the die 21, blank holder force loading, and friction coefficient setting between the die and sheet metal 23. According to the geometric characteristics of the part and the analysis accuracy requirements, a tetrahedral or hexahedral mesh of reasonable density is divided to balance computational efficiency and simulation accuracy. Based on the above settings, nonlinear solution analysis is performed to simulate the forming process of the side panel of the body panel. After the simulation, post-processing analysis is carried out, including extraction of primary and secondary strains of key feature points, forming limit analysis, forming defect identification, evaluation of take-up line fluctuation, and thickness reduction rate analysis. In addition, the simulation results are corrected and optimized through on-site busbar debugging, and the safe size range of the take-up line is finally determined, providing a theoretical basis for the optimization of sheet metal forming process.

[0260] The second step involves using a multi-channel industrial vision inspection module to achieve high-precision image acquisition of the formed sheet metal 23 through synchronous cameras, lenses, and anti-interference devices during the actual production process, ensuring stable image quality and minimal interference at the receiving line.

[0261] Specifically, during the forming analysis process, by identifying the structural features of the body panel sidewall, the distribution of typical defects, and the dimensional fluctuation information of the take-up line, the system assesses the defect risk areas and the dimensional fluctuation sensitive areas of the take-up line during the forming process. Combining finite element simulation results with historical process data, the system focuses on analyzing areas prone to problems such as uneven material flow, strain concentration, and thickness reduction. Based on the above analysis, key image acquisition points are identified, and points on the take-up line corresponding to areas with large stress gradients, abrupt curvature changes, or complex boundaries in the body panel sidewall are prioritized as key image acquisition locations for monitoring in the vision inspection system, in order to achieve early identification of critical defects and dynamic monitoring of dimensional fluctuations.

[0262] The third step involves using the image processing and size calculation module to perform multi-scale edge fusion processing on the acquired images, combining contour detection algorithms / deep learning to accurately extract the material receiving line boundaries, and converting the image data into actual size information with millimeter-level precision through calibration mapping.

[0263] Specifically, the images captured by the camera are converted from their original format to a standard processing format, and grayscale or brightness equalization is performed to eliminate image interference caused by changes in lighting. Gaussian filtering algorithms are applied to reduce noise in the images, removing unstructured noise such as dust and scratches to ensure the stability of subsequent edge detection. Based on the relative positioning of sheet metal 23 in the production cycle, the effective area containing the receiving area in the image is automatically cropped, shortening processing time and improving analysis efficiency. An edge detection algorithm is used to extract the contours of potential receiving lines at different image scales. Image pyramid structure processing technology is introduced to integrate edge response information at different scales. A full-scale edge map containing both coarse and fine boundaries is constructed. Edge response amplitude normalization and non-maximum suppression techniques are used to strengthen the target boundary while suppressing stray edges. Continuous boundaries are extracted using contour detection algorithms, and boundary points are fitted (e.g., least squares line fitting, circular arc fitting, spline interpolation, etc.) to form smooth geometric curves. Alternatively, a lightweight boundary detection deep learning model suitable for industrial inspection (such as HED, DeepEdge, etc.) can be selected. If a deep learning model is used to extract contour curves, existing sheet metal image samples must be used, and training is constructed by manually annotating the true boundaries of the receiving line. The dataset is processed and sample diversity is enhanced (rotation, noise reduction, occlusion reduction). The edge-enhanced image is then input into a deep learning model to obtain a high-confidence pixel-level mask image of the receiving line, automatically filtering out false boundaries and artifacts. Inflection points, peaks and troughs, and maximum offset points in the boundaries are automatically extracted for subsequent dimensioning. A checkerboard array calibration board is used, combined with standard camera calibration methods, to obtain camera intrinsic parameters (focal length, principal point, distortion coefficients, etc.) and extrinsic parameters (pose matrix, etc.). A homography transformation matrix from image pixels to physical space (millimeter-level) coordinates is constructed to achieve unbiased restoration of pixel boundaries to actual boundaries, and radial processing of the image is performed. With tangential distortion correction, the linearity and proportionality accuracy after boundary coordinate transformation are ensured, improving the accuracy of physical measurements. The actual spacing between continuous boundary points is calculated to obtain key dimensional indicators such as the actual length, curvature, and fluctuation amplitude of the take-up line. The boundary vector of the take-up line and the corresponding physical dimension values ​​are superimposed on the image, and abnormal areas are highlighted with red warning boxes. The key point dimension data of the take-up line (maximum deviation, fluctuation range, extreme point, etc.) are output in CSV or JSON format. The processing results are synchronized to the human-computer interaction interface, providing intuitive and readable information such as the take-up line dimension trend chart, stability evaluation indicators, and historical fluctuation trajectory.

[0264] The fourth step involves using a risk identification module to collect real-time dimensional data of key points on the receiving line, performing online fitting of the data points, and then comparing and analyzing them with the threshold range of the receiving line to automatically identify the type of defect risk and trigger an early warning signal. At the same time, fault-tolerant logic is provided to prevent misjudgment.

[0265] Specifically, the system utilizes image processing and size calculation modules to continuously acquire dimensional data (such as edge width and centering error) of key parts of the receiving line, with the data sampling cycle synchronized with the production rhythm. The system pre-introduces threshold ranges for the receiving line and considers reasonable dimensional fluctuation ranges under different operating conditions (such as speed and load). The system compares the real-time collected key point dimensional data with the threshold range under the current state one by one, fits and analyzes the deviation amplitude and trend between the actual size and the threshold range online, improving the stability and anti-interference of the judgment. Based on the analysis results, the system automatically assesses the risk type and sends early warning information.

[0266] Fifth step: Based on the risk identification results, the force value of the hydraulic cylinder in the corresponding area of ​​the first drawing sequence is corrected to achieve local closed-loop control only for the risk area, so as to improve the adjustment accuracy and system response efficiency.

[0267] Furthermore, monitoring is conducted after each adjustment. For example, within 10 to 30 strokes, the receiving line is monitored to see if it returns to the safe range. If not, the edge clamping force is adjusted further.

[0268] In embodiments of the present invention, the terms "first," "second," and "third" are used only for descriptive purposes and should not be construed as indicating or implying relative importance; the term "multiple" refers to two or more unless otherwise expressly defined. The terms "install," "connect," "link," and "fix" should be interpreted broadly. For example, "connect" can be a fixed connection, a detachable connection, or an integral connection; "link" can be a direct connection or an indirect connection through an intermediate medium. Those skilled in the art can understand the specific meaning of the above terms in embodiments of the present invention according to the specific circumstances.

[0269] Furthermore, although the operations are described in a specific order, this should be understood as requiring that such operations be performed in a specific order or sequential order, or requiring that all illustrated operations be performed to achieve the desired result. In certain environments, multitasking and parallel processing may be advantageous. Similarly, although several specific implementation details are included in the above discussion, these should not be construed as limiting the scope of the invention. Certain features described in the context of individual embodiments may also be implemented in combination in a single implementation. Conversely, various features described in the context of a single implementation may also be implemented individually or in any suitable sub-combination in multiple implementations.

[0270] Although the subject matter has been described using language describing specific structural features and / or methodological logic, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or actions described above. Rather, the specific features and actions described above are merely illustrative examples of implementing the claims.

[0271] The above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Those skilled in the art will recognize that various modifications and variations are possible with respect to the embodiments of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the embodiments of the present invention should be included within the protection scope of the embodiments of the present invention.

Claims

1. A sheet metal forming defect risk perception and blank holder force self-feedback control system, characterized in that, include: support; The first punch is mounted on the bracket; Multiple hydraulic devices are connected to the bracket; A blanking ring, which is divided into multiple blanking sections, each of which is connected to the movable end of a hydraulic device, and the sheet material is placed on the blanking ring; The first die is slidably mounted on the bracket and positioned above the sheet metal and the first punch. It can move toward the first punch and, together with the blank holder, presses the outer edge of the sheet metal. Under the combined action of the first punch, it processes the sheet metal into a component. The component includes multiple detection areas, and the detection areas have a mapping relationship with the areas of the sheet metal affected by the blank holder forces of the multiple blank holder sections. An image acquisition component, configured to acquire an image of the outer contour features of the component; An image analysis device is signal-connected to the image acquisition component and the plurality of hydraulic devices. The image analysis device is configured to analyze the defect risk type in the outer contour feature image of each detection area, and generate a pressure compensation command for controlling the hydraulic device of the corresponding pressing section based on the defect risk type and the mapping relationship. After the current forming cycle ends and before the next forming cycle begins, the image analysis device sends the pressure compensation command to the corresponding hydraulic device to adjust the blanking force of the corresponding blanking section. The sheet metal forming defect risk perception and blank holder force self-feedback control system also includes: The testing platform is located on one side of the bracket; A robotic arm, mounted on the inspection platform, is used to grasp the component and transport it to the inspection platform after the current forming cycle is completed. The detection platform serves as the second-order punch, and its surface structure differs from that of the first-order punch. The robotic arm transfers the component formed in the first order to the second order and places it on the detection platform so that the second-order die can descend and complete the processing together with the detection platform. The image acquisition component is mounted on the detection platform via an installation structure; The sheet metal forming defect risk perception and blank holder force self-feedback control system also includes: An electrically controlled slide rail structure includes a guide rail mounted on the mounting structure, a slider meshing with the guide rail, and a first servo motor that drives the slider to move along the guide rail; An angle adjustment device, mounted on the slider, includes a horizontally rotatable gimbal and a second servo motor that drives the gimbal's pitch angle; the image acquisition component is mounted on the gimbal. The product specification database stores the target coordinates and target attitude angles of the image acquisition components corresponding to different product numbers; The parameter mapping module calls the data in the product specification database based on the input product number to generate displacement commands for the electronically controlled slide rail structure and angle commands for the gimbal. The positioning execution unit, connected to the parameter mapping module, controls the first servo motor and the second servo motor to work after receiving the displacement command and the angle command, so as to drive the slider to move to the target coordinates and simultaneously control the gimbal to adjust to the target attitude angle.

2. The sheet metal forming defect risk perception and blank holder force self-feedback control system according to claim 1, characterized in that, The image acquisition component includes: The outer casing is mounted on the gimbal, and its bottom is equipped with a translucent glass. The probe is installed inside the housing via a heat-insulating bracket and can capture images of the outer contour features of the component through the light-transmitting glass. The semiconductor temperature control module includes: A semiconductor cooling chip is disposed on the inner wall of the housing; A temperature sensor is located on the inner wall of the housing; A PID controller is connected to the thermoelectric cooler and the temperature sensor. The PID controller is configured to activate the thermoelectric cooler to cool down when the temperature inside the housing is greater than a first preset temperature, and to activate the thermoelectric cooler to heat up when the temperature inside the housing is less than a second preset temperature.

3. The sheet metal forming defect risk perception and blank holder force self-feedback control system according to claim 1 or 2, characterized in that, The image acquisition component includes a camera and a lens, wherein the camera includes an industrial camera, a security camera, or a combination of both, and the lens includes a fixed-focus lens or a zoom lens.

4. A control method for a sheet metal forming defect risk perception and blank holder force self-feedback control system as described in any one of claims 1 to 3, characterized in that, include: After the current forming cycle is completed, the outer contour feature images of each detection area of ​​the component are acquired by the image acquisition component; The image analysis device analyzes the defect risk type in the outer contour feature image of each detection area and generates a pressure compensation command based on the defect risk type. During the time window from the end of the current forming cycle to the start of the next cycle, the pressure compensation command is sent to the hydraulic device in the corresponding detection area via the industrial bus to adjust the blank holder force for the next forming cycle.

5. The control method of the sheet metal forming defect risk perception and blank holder force self-feedback control system according to claim 4, characterized in that, The step of analyzing the defect risk type in the outer contour feature image of each detection area using an image analysis device includes: A finite element simulation model is constructed based on the property parameters of the sheet metal. Based on the finite element simulation model, a multi-condition simulation experiment was conducted. A first detection point was determined on the sheet metal, and the length L of the take-up line at the first detection point under ideal forming conditions was obtained. 理 The length L of the take-up line at the critical state of wrinkling. 皱 And the length L of the take-up line at the critical cracking state. 裂 ; Based on the L 皱 and the L 理 A safe critical size range for the receiving line is initially set for the first detection point; Based on the critical size range of the take-up line initially determined by the finite element simulation model, and combined with on-site debugging to correct and optimize the simulation results, the critical size of the take-up line [L] was finally determined. min L max ], where L min >L 皱 L max <L 裂 The image acquisition component is calibrated using a checkerboard calibration board to generate a transformation matrix from pixel coordinates to physical space coordinates. The position corresponding to the first detection point is determined in the outer contour feature image and recorded as the second detection point; Calculate the pixel distance from the second detection point to the outer edge of the component; The pixel distance is converted into physical space coordinates using a transformation matrix, and the actual take-up line length L of the second detection point is calculated. 实际 ; L 实际 The critical size range of the receiving line [L] min L max Compare: If L 实际 <L min If so, it is determined that the detection area has a risk of wrinkling; If L 实际 >L max If so, it is determined that there is a risk of cracking in the tested area.

6. The control method of the sheet metal forming defect risk perception and blank holder force self-feedback control system according to claim 5, characterized in that, The steps for constructing a finite element simulation model based on the property parameters of the sheet metal specifically include: Based on the geometry and structural dimensions of the sheet metal, a three-dimensional geometric model including the sheet metal, die, punch, and pressure ring is constructed in the simulation software. Define a material constitutive model for the sheet metal; Define the contact relationship between the sheet metal and the die and punch, and set the coefficient of friction; The blank holder force parameter is applied as a boundary condition to the contact surface between the blank holder ring and the sheet metal; By integrating the three-dimensional geometric model, material constitutive model, contact relationship, and boundary conditions through a finite element solver, a finite element simulation model is generated for simulating the forming process under different blank holder forces and friction coefficients.

Images