A high-speed feeding method for electric release door lock components

Abstract

The application discloses a high-speed feeding method for electric release door lock components, and belongs to the technical field of automobile door lock manufacturing. After detecting that the fixed tool or the clamp jaw is replaced, the tool type information, the allowed receiving direction and the limiting positioning relationship of the fixed tool, and the clamping opening state, the clamping posture state and the clamping action center position of the clamp jaw are read, the receiving constraint domain and the clamping constraint domain are respectively constructed, and the intersection mapping of the receiving constraint domain and the clamping constraint domain is carried out. By establishing the joint feasible domain corresponding to the fixed tool and the clamp jaw after production change, and combining the continuous disc surface state classification, the grabbing result linkage correction and the prediction replenishment control, the feeding equipment can quickly establish a stable control boundary and inhibit disc surface congestion under the production change scene, thereby solving the problems that the existing automatic feeding scheme is slow in stable operation establishment, the risk of misgrabbing and mistransportation is high, and the replenishment and rearrangement states are mismatched.

Description

Technical Field

[0001] This invention relates to the field of automotive door lock manufacturing technology, and more specifically, to a high-speed feeding method for components of electrically released door locks. Background Technology

[0002] The components of electrically released door locks in automobile doors are typically irregular in shape, highly directional, and easily stacked. When using an automated feeding method that combines flexible vibratory feeders, visual recognition, and robotic gripping, although the components on the feeder can be dispersed to a certain extent, problems such as unstable gripping posture of target components, inconsistent feed direction, mismatched receiving direction, and continuous congestion in local areas can still easily occur during actual production changeovers and continuous operation. This can lead to problems such as empty gripping, biased gripping, double gripping, unstable clamping, misjudgment during detection, and cycle fluctuations.

[0003] Existing automated feeding solutions typically only change grippers or fixed fixtures and call preset parameters when replacing different models of parts. They lack a control mechanism to reconstruct the allowable acceptance boundary of the fixed fixture and the gripping boundary of the gripper in a unified manner. They also lack a control mechanism to classify the state evolution of the parts to be fed based on continuous disk images and select differentiated rearrangement and gripping paths based on the classification results. At the same time, existing replenishment control is mainly triggered by fixed material level thresholds, and fails to combine the expected conversion capacity after rearrangement, the proportion of the area of ​​the temporarily ungraspable area, and the back-end transfer cycle time for collaborative judgment. This results in slow establishment of stable operation after production change and easy to continue replenishing materials before the disk is cleared, further increasing the risk of local congestion and misgrabbing. Summary of the Invention

[0004] To address the problems mentioned in the background section, the present invention provides the following technical solution: A high-speed feeding method for components of an electrically released door lock, applied to a feeding device including a flexible vibratory feeder, a vision camera, a robot, a gripper, a fixed fixture, a detection camera, a horizontal conveying mechanism, a hopper, and a belt conveyor mechanism, characterized in that it includes: After detecting that the fixed tooling or the gripper has been replaced, the tooling type information, allowable receiving direction and limit positioning relationship of the fixed tooling, as well as the gripping opening state, gripping posture state and gripping action center position of the gripper are read. The receiving constraint domain and the gripping constraint domain are constructed respectively, and the intersection mapping of the receiving constraint domain and the gripping constraint domain is performed to obtain the joint feasible domain corresponding to the current target model. Based on the joint feasible domain, a single-product operation constraint set corresponding to the current target model is established. Based on at least three frames of disk surface images continuously acquired by the vision camera, the trajectory association of the parts to be loaded on the flexible vibrating disk is performed to determine the pose change trend, discharge direction change trend and adjacent interference state of each part to be loaded. Combined with the single-product operation constraint set, each part to be loaded or disk surface area is classified and judged to form a stable graspable class, a rearrangeable liftable class and a temporarily ungraspable area. Based on the classification result, a unique corresponding control path is output. For the stable graspable class, a direct grasping command is output. For the rearrangeable class, a first rearrangement command is output to control the flexible vibrating disk to execute the first rearrangement vibration sequence. For the temporarily ungraspable area, a second rearrangement command is output to control the flexible vibrating disk to execute a second rearrangement vibration sequence different from the first rearrangement vibration sequence, and to restrict the robot from entering the grasping window of the corresponding area. After the robot completes the grasping and transfer, the validity of the grasp is determined based on the opening and closing displacement changes of the gripper, the gripping drive response changes, the detection results of the detection camera on the parts located on the fixed fixture, and the transfer response results of the horizontal conveying mechanism. When the grasp is determined to be invalid, the front-end material handling parameters and the back-end detection or transfer parameters are adjusted simultaneously, and the corresponding tray area is remarked as a rearrangeable liftable area or a temporarily ungraspable area. After completing the reconstruction of the single-product operation constraint set, the initialization calibration mode is entered. In the initialization calibration mode, at least one initialization grabbing cycle is executed, and the effective grabbing ratio and the area ratio of the temporarily ungrabable area within the initialization grabbing cycle are used to determine whether to switch to continuous feeding mode. In the continuous feeding mode, a predicted remaining quantity of grippable parts is formed based on the number of stable grippable types, the expected number of repositionable liftable types after the first repositioning vibration sequence, the gripping cycle of the robot, and the transfer cycle of the horizontal transport mechanism. When the predicted remaining quantity of grippable parts is lower than a first preset threshold and the area ratio of the temporarily ungripable area is lower than a second preset threshold, the hopper is controlled to replenish material to the flexible vibrating plate via the belt conveyor mechanism. When the predicted remaining quantity of grippable parts is lower than the first preset threshold and the area ratio of the temporarily ungripable area is higher than or equal to the second preset threshold, replenishment is delayed and the second repositioning instruction is executed first.

[0005] Furthermore, the step of establishing a single-product operational constraint set corresponding to the current target model based on the joint feasible domain includes: The intersection of the receiving constraint domain and the clamping constraint domain is mapped to form a joint feasible domain corresponding to the current target model. Based on the joint feasible domain, the consistency screening is performed on the disk surface posture category, disk discharge direction category, and receiving direction category of the parts to be loaded on the flexible vibratory disk; Only the attitude category, exit direction category, and acceptance direction category that simultaneously fall into the joint feasible domain are written into the single-product operation constraint set to form the allowed grab attitude set, allowed exit direction set, and allowed acceptance direction set; And after the fixed fixture or the gripper is replaced, the detection and judgment threshold range and the transfer cycle range are updated synchronously.

[0006] Furthermore, the classification and determination of the parts to be loaded or the disk surface area based on at least three frames of disk surface images includes: Based on at least three frames of disk surface images continuously acquired by the vision camera, the trajectory of each component to be loaded on the flexible vibrating disk is associated. Combined with the gripping opening range of the gripper, the grabbing entry direction of the robot, the allowable receiving direction of the fixed tooling, and the single-product operation constraint set, posture constraint matching value, disk exit direction constraint matching value, receiving direction constraint matching value, and adjacent interference risk value are established for each component to be loaded. When the attitude constraint matching value of a certain component to be loaded is higher than the first threshold, the discharge direction constraint matching value is higher than the second threshold, the receiving direction constraint matching value is higher than the third threshold, and the adjacent interference risk value is lower than the fourth threshold, it is determined to be a stable and graspable type. When a certain component to be loaded does not meet the determination criteria for a stable graspable class, but its posture change trend and position migration trend in adjacent sampling periods converge toward the set of allowed grasping postures, the set of allowed ejection directions, and the set of allowed receiving directions, it is determined to be a rearrangeable liftable class. If the parts to be loaded in a certain area do not meet the criteria for stable graspability in multiple consecutive sampling periods, and the stacking risk and adjacent coupling of the parts to be loaded in that area are continuously higher than the preset threshold, then the area is determined to be a temporarily ungrabable area.

[0007] Furthermore, the first rearrangement instruction is used to control the flexible vibratory feeder to vibrate according to the first rearrangement vibration sequence, which is generated based on the posture convergence direction, feed direction deviation distribution, adjacent interference distribution, and edge clearance distribution of the parts to be loaded within the corresponding area of ​​the rearrangeable lifting class. The first rearranged vibration sequence includes a directional micro-vibration stage along the attitude convergence direction, a stop-vibration and re-shooting stage spaced apart from the directional micro-vibration stage, and a correction vibration stage for correcting the deviation of the disk exit direction. In the stop-vibration and re-shooting stage, the vision camera is called to resample the corresponding area, and the first rearranged vibration sequence is maintained or switched to a direct grasping command based on the resampling result. The second rearrangement instruction is used to control the flexible vibratory feeder to vibrate according to the second rearrangement vibration sequence. The second rearrangement vibration sequence is generated based on the stacking risk distribution, adjacent component coupling distribution, area ratio, and number of consecutive ineffective grasps of the components to be loaded in the corresponding area of ​​the temporarily ungrabable area. The second rearranged vibration sequence includes an enhanced decoupling vibration stage, a diffusion vibration stage, and a boundary evacuation vibration stage; During the second rearrangement vibration sequence, the robot is restricted from entering the grasping window of the corresponding area, and after the second rearrangement vibration sequence is completed, the corresponding area is reclassified.

[0008] Furthermore, the determination of the validity of this crawl includes: After the robot completes a gripping and transfer, based on the opening and closing displacement changes of the gripper, the changes in the gripping drive response, the detection results of the detection camera on the parts located on the fixed fixture, and the transfer response time of the horizontal transport mechanism, the gripping result is determined as an empty grip, an off-center grip, a double grip, an unstable grip, or a valid grip. When the judgment result is empty gripping, biased gripping, double gripping or unstable clamping, at least one of the target screening threshold of the front-end material picking side, the first rearrangement vibration sequence, the second rearrangement vibration sequence and the gripping path of the robot are adjusted simultaneously, as well as at least one of the detection judgment threshold range and the transfer cycle range of the back-end detection or transfer side, and the corresponding tray area where ineffective gripping occurs is remarked as a rearrangeable liftable area or a temporarily ungripable area. When an invalid grasp occurs in the same area within a preset number of consecutive times, the robot is restricted from entering the grasping window of that area, and the control priority for executing the second rearrangement instruction in that area is increased.

[0009] Furthermore, after completing the reconstruction of the single-variety operating constraint set, the system enters the initialization calibration mode; In the initialization calibration mode, the parts to be loaded in the hopper are controlled to be conveyed to the flexible vibratory feeder in a preset initial batch via the belt conveyor mechanism. The flexible vibratory feeder is controlled to execute a preset reference vibration sequence, and an initial stable graspable category ratio, an initial rearrangeable liftable category ratio, and an initial temporarily ungraspable area ratio are formed based on the initial feeder surface image to determine the initial control path corresponding to the current target model. The target screening threshold, the first rearrangement vibration sequence, the second rearrangement vibration sequence, the detection and judgment threshold interval, and the transfer cycle interval are initialized and corrected through at least one initial grasping cycle. In the continuous feeding mode, the predicted amount of grippable parts is formed based on the number of stable grippable types, the expected number of rearranged liftable types after executing the first rearrangement vibration sequence, the gripping cycle of the robot, the transfer cycle of the horizontal transport mechanism, and the area ratio of the temporarily ungripable area. When the predicted remaining amount of grabbable parts is lower than the first preset threshold and the area ratio of the temporarily ungrabable area is lower than the second preset threshold, the hopper is controlled to replenish material to the flexible vibrating plate via the belt conveyor mechanism. When the predicted remaining amount of grabbable parts is lower than the first preset threshold and the area ratio of the temporarily ungrabable area is higher than or equal to the second preset threshold, the replenishment is delayed and the second rearrangement instruction is executed first. After the second rearrangement vibration sequence is completed, the predicted remaining amount of grabbable parts and the area ratio of the temporarily ungrabable area are recalculated, and replenishment is performed when the replenishment conditions are met.

[0010] In summary, the present invention has the following beneficial effects: By establishing a joint feasible domain corresponding to the fixed tooling and grippers after production changeover, and combining continuous tray state classification, gripping result linkage correction, and predictive replenishment control, the feeding equipment can quickly establish a stable control boundary and suppress tray congestion in production changeover scenarios. This solves the problems of slow establishment of stable operation, high risk of misgrabbing and mistransfer, and mismatch between replenishment and rearrangement states in existing automatic feeding solutions. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 This is a schematic diagram of the overall structure of the device of the present invention in a static state; Figure 2 This is a schematic diagram of the flexible vibratory feeder feeding structure of the present invention; Figure 3 This is a schematic diagram of the robot material handling structure of the present invention; Figure 4 This is a schematic diagram of the automatic feeding structure for part image recognition and detection according to the present invention; Figure 5 This is a schematic diagram of the functional module architecture of the control system of the present invention.

[0013] In the picture: 1. Frame; 2. Hopper; 3. Belt conveyor mechanism; 4. Drive motor; 5. Flexible vibratory feeder; 6. Vision camera; 7. Robot; 8. Gripper; 9. Inspection camera; 10. Horizontal handling mechanism; 11. Conveyor belt; 12. Fixed fixture. Detailed Implementation

[0014] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0015] Example 1: The following is in conjunction with the appendix Figures 1-5 The present invention will be described in further detail below.

[0016] Please refer to Figures 1-4 This embodiment provides a feeding device for implementing the above-mentioned high-speed feeding control method for components of an electrically released door lock, including: a frame 1, and a feeding mechanism, a picking mechanism and a detection and transfer mechanism disposed on the frame 1; The feeding mechanism includes a hopper 2, a belt conveyor 3, a drive motor 4, and a flexible vibrating plate 5. One end of the belt conveyor 3 is corresponding to the discharge end of the hopper 2, and the other end is corresponding to the flexible vibrating plate 5. The drive motor 4 is connected to the belt conveyor 3 for driving the components of the electric release door lock located in the hopper 2 to be conveyed to the flexible vibrating plate 5. The flexible vibrating plate 5 is used to disperse and adjust the posture of the conveyed electric release door lock components to form a graspable state. The material handling mechanism includes a vision camera 6, a robot 7, and a gripper 8. The vision camera 6 is set facing the flexible vibrating plate 5 and is used to collect the position and posture information of the electrically released door lock components located on the flexible vibrating plate 5. The robot 7 is set on the frame 1, and the gripper 8 is set at the execution end of the robot 7. The robot 7 is used to grab the electrically released door lock components according to the position and posture information collected by the vision camera 6 and transfer the grabbed electrically released door lock components to the preset inspection station. The inspection and transfer mechanism includes an inspection camera 9, a horizontal transport mechanism 10, a conveyor belt 11, and a fixed fixture 12. The fixed fixture 12 is set at a preset inspection station and is used to position and receive the components of the electric release door lock transferred by the robot 7. The inspection camera 9 is set facing the fixed fixture 12 and is used to perform image recognition and inspection on the components of the electric release door lock located on the fixed fixture 12. The horizontal transport mechanism 10 is set between the fixed fixture 12 and the conveyor belt 11 and is used to transfer the components of the electric release door lock that have been identified and inspected by the inspection camera 9 to the conveyor belt 11 and output them to the target station. The device also includes a control unit, which is connected to the drive motor 4, the flexible vibrating plate 5, the vision camera 6, the robot 7, the detection camera 9, and the horizontal conveying mechanism 10, respectively. In this embodiment, the frame 1 serves as the installation support foundation for the entire machine, supporting the feeding mechanism, the picking mechanism, the detection and transfer mechanism, and the control unit. The feeding mechanism, the picking mechanism, and the detection and transfer mechanism are arranged sequentially along the component transfer direction, so that the components to be fed can operate continuously in the order of feeding, aligning, identifying, picking, detecting, transferring, and output, thereby forming a complete automatic feeding chain.

[0017] In this embodiment, the feeding mechanism includes a hopper 2, a belt conveyor 3, a drive motor 4, and a flexible vibrating plate 5. The hopper 2 is used to temporarily store the parts to be fed. The discharge side of the hopper 2 is corresponding to the input end of the belt conveyor 3, and the output end of the belt conveyor 3 faces the flexible vibrating plate 5. The drive motor 4 is connected to the belt conveyor 3 and is used to drive the belt conveyor 3 to run. In actual operation, the operator first pours the same type of electrically released door lock components into the hopper 2. After the drive motor 4 starts, it drives the belt conveyor 3 to transport the parts to be fed from the hopper 2 to the flexible vibrating plate 5 in batches. In a preferred embodiment, the belt conveyor mechanism 3 may be equipped with an intermittent baffle structure to control the number of parts entering the flexible vibratory feeder 5 at a time, avoiding excessive entry at one time that could cause accumulation, obstruction, or stacking. The flexible vibratory feeder 5 is fixedly installed on the frame 1 and can generate preset vibrations under the control of the control unit, so that the parts entering the feeder surface can be dispersed, flipped, unfolded, and rearranged on the feeder surface, thereby gradually forming a state suitable for subsequent gripping. Compared with the traditional single rigid vibratory feeder, the flexible vibratory feeder 5 is more suitable for irregularly shaped parts, parts with coated surfaces, or parts that are prone to contact scratches, and can reduce the risk of jamming and surface damage.

[0018] The material handling mechanism includes a vision camera 6, a robot 7, and a gripper 8. The vision camera 6 is positioned above or to the side of the flexible vibratory feeder 5, and its field of view covers the effective working area of ​​the flexible vibratory feeder 5. It is used to acquire the distribution, position, and posture of the parts on the feeder surface in real time. The robot 7 is mounted on the frame 1. In this embodiment, a four-axis robot is preferably used to balance gripping speed, motion accuracy, and structural compactness. The gripper 8 is mounted on the execution end of the robot 7 and is used to grip the target parts on the flexible vibratory feeder 5. Preferably, the gripper 8 adopts a modular and replaceable structure to adapt to the differences in size, shape, and gripping position of different types of parts. However, it only performs material handling operations for one type of target part in a single production run. That is, when it is necessary to change the part type, the corresponding gripper 8 and the fixed fixture 12 are replaced, and the control parameters are switched before entering a new single-product operation state, rather than processing multiple different types of parts simultaneously in the same operating cycle.

[0019] The inspection and transfer mechanism includes an inspection camera 9, a horizontal transport mechanism 10, a conveyor belt 11, and a fixed fixture 12. The fixed fixture 12 is set at a preset inspection station, and its function is to receive and position the parts transferred by the robot 7, so that the parts maintain a relatively stable and consistent inspection posture at the inspection station. The inspection camera 9 is set facing the fixed fixture 12, and its front end can be configured with a corresponding lens according to the shape characteristics of the parts, for image recognition and inspection of the parts on the fixed fixture 12, so as to determine whether the current part is the part required for current production, or whether its posture meets the downstream transfer requirements. The horizontal transport mechanism 10 is set between the fixed fixture 12 and the conveyor belt 11. Its function is to transfer the parts on the fixed fixture 12 laterally to the conveyor belt 11 after inspection, and then the conveyor belt 11 transports the parts to the designated target station for subsequent assembly or the next process. In this embodiment, the horizontal transport mechanism 10 can be a horizontal transport cylinder, or other actuators that can achieve the same linear transport function, as long as they can complete the transfer action from the inspection station to the conveyor belt 11.

[0020] The control unit is connected to the drive motor 4, flexible vibratory feeder 5, vision camera 6, robot 7, detection camera 9, and horizontal conveying mechanism 10, respectively, and is used to coordinate and control the actions of the above structures. In this embodiment, the control unit can be implemented by any one or a combination of PLC controller, motion controller, or industrial control host. Its main function is to complete material replenishment control, vibration control, vision acquisition triggering, robot gripping execution, detection triggering, and conveying output control. In specific operation, the control unit first controls the drive motor 4 to drive the belt conveyor mechanism 3 to replenish parts to the flexible vibratory feeder 5, and then controls the flexible vibratory feeder 5 to vibrate, so that the parts on the feeder surface gradually change from a disordered pile state to a dispersed state with a clear directional feature. After the vision camera 6 acquires the image of the feeder surface, the control unit determines the position and posture information of the target parts according to the visual recognition results, and then controls the robot 7 to drive the gripper 8 to grab the parts in the gripping state from the flexible vibratory feeder 5 and transfer them to the fixed fixture 12.

[0021] After a component is placed on the fixed fixture 12, the detection camera 9 performs image recognition detection on the component. If the detection result shows that the component is the component required for current production and its placement meets the preset requirements, the control unit controls the horizontal transport mechanism 10 to move the component from the fixed fixture 12 to the conveyor belt 11, and then the conveyor belt 11 transports it to the required position. If the detection result does not meet the requirements, the control unit can control the horizontal transport mechanism 10 not to perform the transfer action, or guide the component to a preset unqualified processing path. In a preferred embodiment, when the vision camera 6 detects that the number of components in the flexible vibrating plate 5 that are in a graspable state has decreased to below a preset number, the control unit controls the drive motor 4 to start again, so that the belt conveyor mechanism 3 replenishes the components in the hopper 2 into the flexible vibrating plate 5 in batches, so as to ensure that the robot 7 has a graspable target, thereby maintaining the continuous feeding capacity of the whole machine.

[0022] The working principle of this embodiment can be summarized as follows: First, the raw bulk materials are quantitatively supplied through the hopper 2 and the belt conveyor mechanism 3. Then, the flexible vibrating plate 5 adjusts the state and disperses the parts spatially, so that the originally disordered parts are gradually formed into a plate surface state that can be visually recognized and grasped by the robot. Subsequently, the vision camera 6 collects images of the parts on the flexible vibrating plate 5, and the robot 7 grasps the target parts according to the recognition results and sends them to the fixed fixture 12. Next, the detection camera 9 is used to re-identify and confirm the parts on the fixture to avoid mixing parts with similar shapes. Finally, the horizontal transport mechanism 10 and the conveyor belt 11 transport the identified parts to the target position, thereby completing the fully automatic feeding process from bulk material entry to qualified part output. In this process, the flexible vibrating plate 5 is responsible for organizing the state of the parts, the vision camera 6 is responsible for plate surface recognition and positioning, the robot 7 and the gripper 8 are responsible for grasping and transporting, the fixed fixture 12 and the detection camera 9 are responsible for detection and confirmation, and the horizontal transport mechanism 10 and the conveyor belt 11 are responsible for output transfer. Each structure has a clear division of labor and is continuously connected.

[0023] Taking a component of an electric release door lock as an example, in actual operation, the operator first pours the component into the hopper 2. After the control unit is activated, the belt conveyor 3 feeds the components into the flexible vibrating plate 5 in batches. Under the vibration of the flexible vibrating plate 5, the components gradually change from an overlapping, flipped, or irregularly placed state to a state suitable for gripping. After the vision camera 6 takes a picture of the plate surface, the robot 7 selects the components that meet the gripping requirements, grips them with the gripper 8, and places them on the fixed fixture 12. The detection camera 9 identifies the components on the fixed fixture 12. After confirming that they are the current target components, the horizontal transport mechanism 10 pushes them onto the conveyor belt 11, and the conveyor belt 11 then sends them to the designated position. If the vision camera 6 detects that the number of grippable components in the flexible vibrating plate 5 has decreased, the control unit activates the replenishment action, so that the components in the hopper 2 re-enter the flexible vibrating plate 5, thereby achieving continuous cyclic operation.

[0024] This equipment platform provides the foundation for the implementation of the control methods described later. The combination of its hopper 2, belt conveyor mechanism 3, and flexible vibratory feeder 5 enables continuous supply and status management of parts, reducing the problems of easy jamming and scratching under traditional rigid vibration methods. Furthermore, through the cooperation of vision camera 6 and robot 7, it can achieve automatic positioning and gripping of parts on the feeder surface, improving the feeding speed and gripping accuracy. Thirdly, through the re-identification and confirmation by fixed fixture 12 and detection camera 9, the risk of mixing and misfeeding of parts with similar shapes can be effectively reduced. By changing the gripper 8 and fixed fixture 12, it can adapt to the switching needs of different models of electrically released door lock components, thereby improving the applicability and production changeover convenience of the equipment in automotive parts production.

[0025] It should be noted that the above embodiments are only a specific description of the device. Without departing from the overall concept of the present invention, the specific type of robot 7, the specific structural form of horizontal conveying mechanism 10, the clamping structure of gripper 8, and the lens configuration of detection camera 9 can all be adjusted according to the actual component types.

[0026] Example 2 A high-speed feeding method for components of an electrically released door lock, such as Figures 1-5 As shown, after detecting that the fixed fixture 12 or the gripper 8 has been replaced, the fixture type information, allowable bearing direction and limit positioning relationship of the fixed fixture 12, as well as the gripping opening state, gripping posture state and gripping action center position of the gripper 8 are read. The bearing constraint domain and the gripping constraint domain are constructed respectively, and the intersection mapping of the bearing constraint domain and the gripping constraint domain is performed to obtain the joint feasible domain corresponding to the current target model. Based on the joint feasible domain, a single-product operation constraint set corresponding to the current target model is established. Based on at least three frames of disk surface images continuously acquired by the vision camera 6, the trajectory association of the parts to be loaded on the flexible vibrating disk 5 is performed to determine the pose change trend, discharge direction change trend and adjacent interference state of each part to be loaded. Combined with the single-product operation constraint set, each part to be loaded or disk surface area is classified and judged to form a stable graspable class, a rearrangeable lifting class and a temporarily ungraspable area. Based on the classification and judgment results, output a unique corresponding control path. For stable graspable classes, output a direct grasping command. For rearrangeable and liftable classes, output a first rearrangement command to control the flexible vibrating disk 5 to execute the first rearrangement vibration sequence. For areas that are temporarily ungraspable, output a second rearrangement command to control the flexible vibrating disk 5 to execute a second rearrangement vibration sequence that is different from the first rearrangement vibration sequence, and restrict the robot 7 from entering the grasping window of the corresponding area. After robot 7 completes the gripping and transfer, the validity of the gripping is determined based on the opening and closing displacement changes of gripper 8, the gripping drive response changes, the detection results of the detection camera 9 on the parts located on the fixed fixture 12, and the transfer response results of the horizontal conveying mechanism 10. When the gripping is determined to be invalid, the front-end material handling parameters and the back-end detection or transfer parameters are adjusted simultaneously, and the corresponding tray area is remarked as a rearrangeable liftable area or a temporarily ungripable area. After completing the reconstruction of the single-product operation constraint set, the system enters the initialization calibration mode. In the initialization calibration mode, at least one initialization grabbing cycle is executed, and the system determines whether to switch to the continuous feeding mode based on the effective grabbing ratio and the area ratio of the temporarily ungrabable area within the initialization grabbing cycle. In continuous feeding mode, based on the number of stable graspable items, the expected number of repositionable items to be converted after the first repositioning vibration sequence, the grasping cycle of robot 7, and the transfer cycle of horizontal conveying mechanism 10, a predicted remaining amount of graspable parts is formed. When the predicted remaining amount of graspable parts is lower than the first preset threshold and the area of ​​the temporarily ungraspable area is lower than the second preset threshold, the material bin 2 is controlled to replenish material to the flexible vibrating plate 5 via the belt conveyor mechanism 3. When the predicted remaining amount of graspable parts is lower than the first preset threshold and the area of ​​the temporarily ungraspable area is higher than or equal to the second preset threshold, the replenishment is delayed and the second repositioning instruction is executed first. Based on the joint feasible domain, establish a single-product operational constraint set corresponding to the current target model, including: The intersection of the receiving constraint domain and the clamping constraint domain is mapped to form a joint feasible domain corresponding to the current target model. Based on the joint feasible domain, the consistency screening is performed on the disk surface posture category, disk discharge direction category, and receiving direction category of the parts to be loaded on the flexible vibratory disk 5. Only the attitude category, exit direction category, and acceptance direction category that simultaneously fall into the joint feasible domain are written into the single-product operation constraint set to form the allowed grab attitude set, allowed exit direction set, and allowed acceptance direction set; And after the fixed fixture 12 or the gripper 8 is replaced, the detection judgment threshold range and the transfer cycle range are updated synchronously.

[0027] In this embodiment, the "drive lever A" from an electric release door lock is used as the target feeding component, and a "drive lever B" with a similar outer contour but different installation direction requirements is selected as an interference sample to verify the actual effect of the control system in high-speed feeding, suppression of mis-grabbing, and prevention of mixed materials. The feeding system of the electric release door lock used in the test platform includes a frame 1, a hopper 2, a belt conveyor mechanism 3, a drive motor 4, a flexible vibratory feeder 5, a vision camera 6, a robot 7, grippers 8, a detection camera 9, a horizontal transport mechanism 10, a conveyor belt 11, a fixed fixture 12, and a control unit. Before the test, the fixed fixture 12 and grippers 8 corresponding to the original production model are removed and replaced with fixed fixture 12 and grippers 8 corresponding to drive lever A. The fixed fixture 12 defines the landing reference plane and allowable bearing direction of drive lever A, and the grippers 8 define the gripping opening range and gripping center of the robot 7 when gripping drive lever A. After replacement, the control unit activates the tooling constraint establishment unit, reads the tooling type information, allowed receiving direction, and limit positioning relationship of the fixed tooling 12, and simultaneously reads the clamping opening state and clamping posture state of the gripper 8, and establishes a single-product operation constraint set corresponding to the current target model. Subsequently, 1200 drive levers A are added to the hopper 2, and 24 drive levers B are artificially mixed in to simulate the occasional mixing scenario in the production site of similar parts; the belt conveyor mechanism 3 conveys parts to the flexible vibratory feeder 5 in batches of 25. Under the control of the control unit, the flexible vibratory feeder 5 first performs a basic vibration for 1.5 s, and then enters the continuous rearrangement operation state; the vision camera 6 continuously acquires 3 frames of feeder surface images at a sampling interval of 120 ms. The feeder surface state evolution unit calculates the posture constraint matching value, feed direction constraint matching value, receiving direction constraint matching value, and adjacent interference risk value based on the posture changes, position migrations, edge clearance changes, and adjacent coupling changes of each part in the 3 frames of images. In this embodiment, the threshold for determining a stable graspable class is set as follows: the attitude constraint matching value is greater than 0.8, the output direction constraint matching value is greater than 0.75, the receiving direction constraint matching value is greater than 0.7, and the adjacent interference risk value is less than 0.3. When the target component meets the above conditions, the control path decision unit only outputs a direct grasping command, and the robot 7 immediately drives the gripper 8 to grasp and transfer the component to the fixed fixture 12. When the target component does not meet the stable graspable conditions, but its attitude change trend and position migration trend converge toward the set of allowable grasping attitudes and the set of allowable receiving directions, the control path decision unit only outputs a first rearrangement command, and the flexible vibrating disk 5 executes a first rearrangement vibration sequence consisting of directional micro-vibration, vibration stop and re-beat, and correction vibration. When a local area cannot meet the grasping conditions for multiple consecutive sampling periods and the stacking risk remains high, the control path decision unit only outputs a second rearrangement command, and the flexible vibrating disk 5 executes a second rearrangement vibration sequence consisting of enhanced decoupling vibration, diffusion vibration, and boundary evacuation vibration, and restricts the robot 7 from entering the area for grasping.After robot 7 completes the gripping and places the part onto the fixed fixture 12, the detection camera 9 performs secondary recognition on the part, including the part's contour features, the direction of key holes, and its placement status on the fixed fixture 12. After successful recognition, the horizontal transport mechanism 10 transfers the part to the conveyor belt 11 and then outputs it to the target station. The gripping result linkage write-back unit synchronously records the opening and closing displacement changes of the gripper 8, the gripping drive response changes, the recognition results of the detection camera 9, and the transfer response time of the horizontal transport mechanism 10. Each gripping result is judged as an empty grip, an off-center grip, a double grip, an unstable grip, or a valid grip. When an invalid grip occurs, the relevant parameters in the target screening threshold, the first rearrangement vibration sequence, the second rearrangement vibration sequence, the robot 7 gripping path, the detection judgment threshold range, and the transfer cycle interval are simultaneously corrected. The remaining quantity prediction and replenishment unit predicts the remaining quantity of grabbable parts in the next replenishment cycle based on the number of stable grabbable items, the expected conversion quantity of rearrangeable and liftable items, the grabbing cycle of robot 7, the transfer cycle of horizontal transport mechanism 10, and the area ratio of temporarily ungrabable areas. If the predicted remaining quantity is insufficient and the congestion of the pallet is within the allowable range, the belt conveyor mechanism 3 is activated for replenishment. If the area of ​​temporarily ungrabable areas on the pallet is too large, the second rearrangement vibration is executed first, and replenishment is delayed.

[0028] To verify the effectiveness of the control system of this invention, four groups of test subjects were set up: Invention Example, Comparative Example 1, Comparative Example 2, and Comparative Example 3. The high-speed feeding control method used in the Invention Example includes joint feasible domain establishment, trajectory association classification based on at least three frames of disk surface images, unique control path output, gripping validity determination and linkage correction, initialization calibration, and predictive feeding. Comparative Example 1 uses a conventional control method of "only calling preset parameters after tooling change," without establishing joint constraints between the fixed tooling 12 and the gripper 8. Comparative Example 2 uses a control method of "direct gripping after single-frame visual recognition," without classifying disk surface state evolution. Comparative Example 3 uses a control method of "only local correction after gripping and feeding at low material level thresholds," without front-to-back linkage write-back or predictive feeding. Each test group ran continuously for 60 minutes, repeated 3 times, and the average value was used as the data in the table below.

[0029] Table 1: Comparison of high-speed feeding performance of components for electrically released door locks under different control strategies Average stable crawlable class percentage / % 72.8 61.4 53.9 58.6 Average effective crawl rate / % 98.7 94.3 89.6 92.8 Missed capture rate / % 0.4 1.9 4.1 2.3 Off-target capture rate / % 0.5 1.7 3.0 2.0 Double capture rate / % 0.2 0.8 1.6 1.0 Mixing error loading rate / % 0.08 0.42 0.76 0.35 Average feeding cycle time / piece per minute 52.6 48.7 45.4 47.1 Beat fluctuation coefficient / % 3.2 7.9 11.4 8.8 Stable production time after production change / s 165 438 526 372 Average area percentage of areas temporarily unscrapable / % 10.6 18.7 24.3 19.5 Number of manual interventions per 60-minute continuous operation 0 2 5 3 First inspection pass rate / % 99.2 97.8 96.5 97.1 The data in the table shows that the invention significantly outperforms the three comparative examples in all key indicators. This advantage is not a localized improvement brought about by a single method, but rather a systematic effect formed by "tooling constraint establishment, tray state evolution, unique control path decision-making, linked write-back of grasping results, and surplus prediction and replenishment." First, in terms of the average stable graspable target percentage, the invention reaches 72.8%, significantly higher than Comparative Example 1's 61.4%, Comparative Example 2's 53.9%, and Comparative Example 3's 58.6%. This indicates that the invention does not simply use visual recognition results as the grasping basis, but rather, under the constraints jointly determined by the fixed tooling 12 and the gripper 8, it further filters the tray targets, allowing only targets that simultaneously meet the requirements of grasping posture, tray exit direction, and receiving direction to enter the direct grasping path. Therefore, it can significantly increase the proportion of truly executable graspable targets.

[0030] Although Comparative Example 1 also called parameters after tooling switching, because the joint feasible domain between fixed tooling 12 and gripper 8 was not established, some targets that "visually appear to be graspable but are actually unstable after being placed in fixed tooling 12" were incorrectly included in the grasping range, so the proportion of stable graspable targets was significantly low.

[0031] Comparative Example 2, due to its use of single-frame recognition for direct capture, included a large number of temporarily revealed but unstable targets on the disk surface in the capture window, resulting in the lowest proportion of stable and captureable targets. Secondly, considering the combined data of average effective capture rate, empty capture rate, biased capture rate, and double capture rate, the invention example achieved 98.7%, 0.4%, 0.5%, and 0.2% respectively, outperforming the other groups.

[0032] This data directly reflects the inventive effect of this invention in that "the classification result uniquely determines the control path": For stable graspable classes, only direct grasp instructions are output; for rearranged and upliftable classes, only the first rearrangement instruction is output; for temporarily ungraspable areas, only the second rearrangement instruction is output and robot 7 is restricted from entering the grasping window. As a result, robot 7 no longer blindly attempts to grasp boundary state targets, but only performs grasping after the control system confirms that it meets multiple constraints, thus significantly suppressing empty grasping, off-center grasping, and double grasping.

[0033] The data in Comparative Example 2 is the worst, precisely because it lacks constraints on the evolution of the board state and classification, and the grasping action is too dependent on the instantaneous image results.

[0034] Furthermore, considering the three indicators—mixed material misfeeding rate, first-pass yield, and time to stable production after production changeover—the invention's results are 0.08%, 99.2%, and 165 seconds, respectively, significantly better than the other groups. The reason the mixed material misfeeding rate can be reduced to 0.08% is that this invention does not treat the detection camera 9 merely as a simple downstream verification device, but rather incorporates its detection results into the grasping result linkage write-back link. Once the detection finds that the current component does not conform to the target model or the placement direction is incorrect, the system will simultaneously write back the parameters of the front-end material handling side and the back-end detection and transfer side, thereby enabling stricter control over similar risk areas in subsequent cycles. The shortest time to stable production after production changeover demonstrates the practical value of the tooling constraint establishment unit of this invention. After the fixed tooling 12 and gripper 8 are replaced, the control system can quickly establish a single-product operating constraint set based on tooling type information, allowed receiving direction, and clamping status, rather than relying on repeated manual trial grasping and debugging. Therefore, it can restore stable operation more quickly in production changeover scenarios.

[0035] Finally, the average feeding cycle time, cycle time fluctuation coefficient, average area ratio of temporarily ungraspable zones, and number of manual interventions in 60 minutes of continuous operation demonstrate that this invention also has significant advantages in continuous high-speed operation. The invention example achieves an average feeding cycle time of 52.6 pieces / min, a cycle time fluctuation coefficient of only 3.2%, an average area ratio of temporarily ungraspable zones of only 10.6%, and zero manual interventions in 60 minutes of continuous operation. This indicates that the invention not only grasps accurately but also grasps stably over a long period. The fundamental reason is that the inventory prediction replenishment unit does not replenish only when the number of pieces in the tray is low. Instead, it comprehensively considers the stable number of graspable items, the expected conversion quantity of items that can be improved through rearrangement, the feeding cycle time, the transfer cycle time, and the area ratio of temporarily ungraspable zones to determine when to replenish and when to rearrange before replenishing. Therefore, it avoids the stacking expansion and cycle time deterioration caused by continuing to replenish before the tray is cleared.

[0036] While Comparative Example 3 retains some recognition and detection functions, it lacks front-end and back-end linkage write-back and predictive material replenishment. It can only rely on fixed thresholds to trigger material replenishment and perform local corrections, resulting in a significantly lower cycle time and fewer manual interventions compared to the Invention Example. Based on the above data, it can be seen that this invention is not a simple parameter optimization of existing flexible vibration plus vision gripping schemes. Instead, it establishes a closed-loop control mechanism involving the fixed fixture 12, gripper 8, flexible vibratory feeder 5, vision camera 6, robot 7, detection camera 9, and horizontal transport mechanism 10. This mechanism simultaneously improves gripping accuracy, mis-grip suppression, changeover recovery speed, material mixing prevention, and continuous operation stability during high-speed material feeding.

[0037] Example 3 like Figures 1-5 As shown, the classification and determination of the parts to be loaded or the area of ​​the disk surface are based on at least three frames of disk surface images, including: Based on at least three frames of disk surface images continuously acquired by vision camera 6, the trajectory association of each component to be loaded on the flexible vibrating disk 5 is performed. Combined with the gripping opening range of gripper 8, the grasping entry direction of robot 7, the allowable receiving direction of fixed tooling 12, and the single-product operation constraint set, posture constraint matching value, disk exit direction constraint matching value, receiving direction constraint matching value, and adjacent interference risk value are established for each component to be loaded. When the attitude constraint matching value of a certain component to be loaded is higher than the first threshold, the discharge direction constraint matching value is higher than the second threshold, the receiving direction constraint matching value is higher than the third threshold, and the adjacent interference risk value is lower than the fourth threshold, it is determined to be a stable and graspable type. When a certain component to be loaded does not meet the criteria for stable grasping, but its attitude change trend and position migration trend in adjacent sampling periods converge toward the set of allowed grasping attitudes, the set of allowed ejection directions, and the set of allowed receiving directions, it is judged as a rearrangeable lifting component. If the parts to be loaded in a certain area do not meet the criteria for stable graspability in multiple consecutive sampling periods, and the stacking risk and adjacent coupling of the parts to be loaded in that area are continuously higher than the preset threshold, then the area is determined to be a temporarily ungrabable area. In this embodiment, the "drive lever A" in the electric release door lock of an automobile is used as the target feeding component, and the "drive lever B" with a similar shape but different hole orientation is selected as the interference component to verify the actual role of the disk state evolution classification and grasping result linkage write-back mechanism in the high-speed feeding scenario. The test platform adopts a high-speed feeding method for the components of the electric release door lock. The device includes a frame 1, a hopper 2, a belt conveyor mechanism 3, a drive motor 4, a flexible vibrating plate 5, a vision camera 6, a robot 7, a gripper 8, a detection camera 9, a horizontal transport mechanism 10, a conveyor belt 11, a fixed fixture 12, and a control unit.

[0038] Before the test, clamping jaws 8 and fixing fixtures 12 matching the drive lever A were installed. The clamping opening range of the clamping jaws 8 was set to 16 mm to 21 mm, and the clamping center was located 1.8 mm off-center from the geometric center of the drive lever A. The allowable bearing direction of the fixing fixture 12 was set so that the mounting hole faces the inspection camera 9 and the flange faces the horizontal conveying mechanism 10. Subsequently, 1000 drive lever A pieces were added to the hopper 2, along with 20 drive lever B pieces, to simulate occasional mixing situations in the production of similar parts.

[0039] After the control unit is started, the drive motor 4 drives the belt conveyor mechanism 3 to feed materials to the flexible vibratory feeder 5 at a rate of 20 pieces per batch. The flexible vibratory feeder 5 first performs a basic vibration for 1.2 seconds, and then enters a continuous rearrangement state. The vision camera 6 continuously acquires 3 frames of feeder surface images at a sampling interval of 120 ms. Based on the posture changes, position migrations, adjacent component spacing changes, and edge clearance changes of each component in the 3 frames of images, the feeder surface state evolution unit establishes posture constraint matching values, feed-out direction constraint matching values, receiving direction constraint matching values, and adjacent interference risk values ​​for each component to be fed. In this embodiment, the threshold values ​​for posture constraint matching values ​​are set to 0.8, feed-out direction constraint matching values ​​are set to 0.75, receiving direction constraint matching values ​​are set to 0.7, and adjacent interference risk values ​​are set to 0.3.

[0040] When a component meets the condition that the first three matching values ​​are all higher than the corresponding thresholds and the adjacent interference risk value is lower than 0.3, it is marked as a stable graspable class and only the control path decision unit is allowed to output direct grasping instructions; when it does not currently meet the stable graspable class judgment conditions, but the attitude change direction and position migration direction in adjacent sampling periods are obviously converging towards the set of allowed grasping attitudes, the set of allowed dispensing directions, and the set of allowed receiving directions, it is marked as a rearrangement-promotable class and only the first rearrangement instruction is allowed to be output. When a local area fails to meet the criteria for stable graspability for five consecutive sampling cycles, and the stacking risk and adjacent coupling of components within that area remain at a high level, it is marked as a temporarily ungraspable area. Only the second rearrangement command is allowed, and robot 7 is restricted from entering this area to perform grasping. For rearrangeable liftable targets, the flexible vibratory feeder 5 executes a first rearrangement vibration sequence, including a 0.25 s directional micro-vibration stage, a 0.10 s stop-vibration and re-shoot stage, and a 0.18 s correction vibration stage. For temporarily ungraspable areas, the flexible vibratory feeder 5 executes a second rearrangement vibration sequence, including a 0.30 s enhanced decoupling vibration stage, a 0.22 s diffusion vibration stage, and a 0.15 s boundary evacuation vibration stage. After robot 7 grasps a stable graspable target and places it on the fixed fixture 12, the detection camera 9 re-identifies the component contour, hole orientation, and placement status. The horizontal transport mechanism 10 records the response time from the moment the detection is passed to the completion of the transport action.

[0041] The grasping result linkage write-back unit synchronously collects the opening and closing displacement changes of gripper 8, the gripping drive response changes, the recognition results of detection camera 9, and the transfer response results of horizontal conveying mechanism 10, and judges each grasp as empty grasp, biased grasp, double grasp, unstable grip, or effective grasp. When an ineffective grasp occurs, the system not only corrects the target screening threshold, the first rearrangement vibration sequence, the second rearrangement vibration sequence, and the robot 7 grasping path on the front-end material picking side, but also synchronously corrects the back-end detection judgment threshold range and the transfer cycle interval, and re-marks the abnormal tray area as the area corresponding to the rearrangement liftable type or the area corresponding to the temporarily ungraspable area.

[0042] To verify the effectiveness of this invention, four groups of test subjects were set up. The invention example employed a complete classification plus linked write-back mechanism; Comparative Example 1 used a single-frame recognition followed by direct capture, without performing three consecutive frames of state evolution classification; Comparative Example 2 used three-frame recognition and classification, but the classification result did not uniquely determine the control path, allowing for trial capture while classifying; Comparative Example 3 used three-frame classification and a unique control path, but after capture, only local corrections were made to the capture path, without linked write-back of the detection judgment threshold interval and the transport cycle interval. Each group ran continuously for 60 minutes, repeated three times, and the average value was taken as the test result.

[0043] Table 2: Comparison of Control Mechanism Test Results Average percentage of stable crawlable categories / % 70.4 55.8 61.7 66.2 Rearrangement can improve the percentage of classes transitioning to stable classes / % 46.9 18.5 29.4 41.1 Average area percentage of areas temporarily unscrapable / % 9.8 23.6 18.7 13.9 Effective crawl rate / % 98.4 90.7 94.2 96.1 Missed capture rate / % 0.5 3.8 1.9 1.1 Off-target capture rate / % 0.6 2.9 1.8 1.2 Double capture rate / % 0.2 1.4 0.9 0.5 Clamping instability rate / % 0.3 1.2 0.8 1.1 Mixed material mistransfer rate / % 0.07 0.68 0.31 0.18 Average feeding cycle time / piece per minute 52.1 45.0 48.2 49.4 Beat fluctuation coefficient / % 3.5 11.2 7.6 5.9 Number of manual interventions per 60 consecutive minutes 0 5 3 2 Regional repeat anomaly incidence rate / % 2.1 14.8 8.9 6.4 As shown in Table 2, the invention outperforms the three comparative examples in key indicators such as the average proportion of stable graspable classes, the proportion of rearranged upgradable classes transitioning to stable classes, effective grasping rate, and average feeding cycle time. Furthermore, it significantly lowers failure-related indicators such as empty grasping rate, biased grasping rate, double grasping rate, clamping instability rate, mixed material mistransfer rate, cycle time fluctuation coefficient, and number of manual interventions. This indicates that the disc state evolution classification mechanism and the grasping result linkage write-back mechanism are not simply additional measures in conventional visual grasping systems, but rather form a closed-loop control chain with significant technical effects. Firstly, considering the average proportion of stable graspable classes and the proportion of rearranged upgradable classes transitioning to stable classes, the invention achieves 70.4% and 46.9% respectively, significantly higher than Comparative Example 1's 55.8% and 18.5%, and also higher than Comparative Examples 2 and 3. This demonstrates that trajectory association and constraint matching judgment under three consecutive frames can more accurately distinguish between targets that are "currently graspable" and those "grabable after rearrangement" than single-frame recognition. Comparative Example 1 relies solely on single-frame recognition, failing to identify whether the target's posture is converging towards the set of allowed grasping postures and allowed receiving directions. Consequently, a large number of boundary-state targets are directly included in the grasping window, resulting in a low proportion of truly stable graspable targets. Although Comparative Example 2 introduces three-frame classification, the classification results do not uniquely determine the control path, causing the system to still attempt to grasp some rearranged upliftable targets, resulting in insufficient rearrangement transformation. Furthermore, the average area ratio of temporarily ungraspable regions in the invention example is only 9.8%, far lower than 23.6% in Comparative Example 1, 18.7% in Comparative Example 2, and 13.9% in Comparative Example 3. This indicates that the invention does not passively wait for the board to improve naturally, but actively promotes the evolution of the board from a disordered state to a graspable state by dividing the board targets into stable graspable, rearranged upliftable, and temporarily ungraspable regions, and corresponding to three unique control branches: direct grasp, first rearrangement, and second rearrangement. Furthermore, considering the combined data of effective capture rate, empty capture rate, biased capture rate, double capture rate, and gripping instability rate, the invention example achieved 98.4%, 0.5%, 0.6%, 0.2%, and 0.3% respectively, which is significantly better than the three comparison examples. This result directly reflects the synergistic advantage of "front-end classification decision-making plus back-end linkage write-back".

[0044] Although Comparative Example 3 retains the three-frame classification and unique control path, it only makes local corrections to the grasping path after grasping, without simultaneously correcting the detection and judgment threshold range and the transfer cycle time range. Therefore, the clamping instability rate still reaches 1.1%, indicating that only correcting the front end without correcting the back end cannot suppress anomalies at the overall machine level. Regarding the mixed material mistransfer rate, the invention example is only 0.07%, while Comparative Example 1 reaches 0.68%. This shows that the recognition result of the detection camera 9 can only truly form a continuous interception capability for mixed parts after being linked back to the front-end classification and back-end transfer judgment logic, rather than making an isolated judgment only once in the current cycle. In terms of average feeding cycle time and cycle time fluctuation coefficient, the invention example is 52.1 pieces per minute and 3.5% respectively, which is better than Comparative Example 1's 45.0 pieces per minute and 11.2%, and also better than the other comparative examples. This shows that the present invention does not sacrifice cycle time by adding classification and write-back steps. On the contrary, it makes the continuous feeding process more stable by reducing repeated abnormal areas, reducing manual intervention, and compressing the area of ​​temporarily ungraspable areas. Finally, the regional repetition anomaly rate in the invention example was only 2.1%, far lower than 14.8% in Comparative Example 1, 8.9% in Comparative Example 2, and 6.4% in Comparative Example 3. This is the direct effect of "re-marking and linking the corresponding disk area where ineffective grasping occurs." In other words, the present invention does not patch each anomaly separately, but rather reintegrates the anomaly area into the disk state evolution and control path decision chain, so that the problem area is more strictly classified, rearranged, and grasped in subsequent cycles. As can be seen from the data in Table 2, the technical solution corresponding to Example 3 simultaneously improved the disk's effective target formation capability, grasping accuracy, material mixing prevention capability, and continuous operation stability in actual experiments. This simultaneous improvement of multiple indicators is not a result that can be naturally obtained by conventional single-frame recognition and grasping schemes or single-point correction schemes, but is based on the combined effect of three key distinguishing features: continuous three-frame state evolution classification, unique control path output, and cross-segment linkage write-back.

[0045] Example 4 like Figures 1-5 As shown, the first rearrangement vibration sequence is generated based on the attitude convergence direction, discharge direction deviation distribution, adjacent interference distribution and edge clearance distribution of the parts to be loaded in the corresponding area of ​​the rearrangeable lifting class. The first rearranged vibration sequence includes a directional micro-vibration stage along the attitude convergence direction, a stop-vibration and re-shooting stage set at intervals with the directional micro-vibration stage, and a correction vibration stage to correct the deviation of the disk exit direction. In the stop-vibration and re-shooting stage, the vision camera 6 is called to resample the corresponding area, and the result of the resampling is used to determine whether to maintain the first rearranged vibration sequence or switch to the direct grasping command. The second rearrangement vibration sequence is generated based on the stacking risk distribution, adjacent component coupling distribution, area ratio, and number of consecutive ineffective grasps of the components to be loaded in the corresponding area of ​​the temporarily ungrabable area. The second rearranged vibration sequence includes an enhanced decoupling vibration stage, a diffusion vibration stage, and a boundary evacuation vibration stage. During the execution of the second rearranged vibration sequence, robot 7 is restricted from entering the grasping window of the corresponding area, and the corresponding area is reclassified after the execution of the second rearranged vibration sequence. The validity of this data crawl will be determined, including: After robot 7 completes a gripping and transfer, based on the opening and closing displacement changes of gripper 8, the changes in gripping drive response, the detection results of the detection camera 9 on the parts located on the fixed fixture 12, and the transfer response time of the horizontal transport mechanism 10, the gripping result is judged as empty gripping, biased gripping, double gripping, unstable gripping, or effective gripping. When the judgment result is empty gripping, biased gripping, double gripping or unstable clamping, at least one of the target screening threshold, first rearrangement vibration sequence, second rearrangement vibration sequence and gripping path of robot 7 on the front-end material picking side is adjusted simultaneously, as well as at least one of the detection judgment threshold range and transfer cycle range on the back-end detection or transfer side, and the corresponding plate area where ineffective gripping occurs is remarked as a rearrangeable liftable area or a temporarily ungripable area. When an invalid grasp occurs in the same area within a preset number of consecutive times, the robot 7 is restricted from entering the grasping window of that area, and the control priority for executing the second rearrangement instruction in that area is increased. In this embodiment, to verify the actual effect of the control path decision unit in generating the first and second rearranged vibration sequences for different disc states, the "locking lever A" in the automotive electric release door lock was selected as the target loading component, and the "locking lever B" with similar contour dimensions but different edge flange directions was selected as the interference component. The test platform adopts a high-speed loading method and is equipped with a control system. Before the test, the gripper 8 and the fixing fixture 12 that match the locking lever A were installed on the equipment. The effective clamping opening range of the gripper 8 was set to 14 mm to 19 mm, and the fixing fixture 12 limited the receiving direction and stop position of the locking lever A when it enters the detection station.

[0046] To amplify the differences during the rearrangement process on the disc, 1200 locking levers A and 24 locking levers B were added to hopper 2. Simultaneously, the initial material density was artificially increased to make the initial components entering the flexible vibratory feeder 5 more prone to local overlap, edge aggregation, and directional confusion. Drive motor 4 drives belt conveyor 3 to deliver components to the flexible vibratory feeder 5 in batches of 24. The flexible vibratory feeder 5 first performs a basic vibration for 1.0 s, followed by continuous image acquisition of the disc surface by vision camera 6 at 100 ms sampling intervals.

[0047] The disk state evolution unit extracts the orientation convergence direction, disk exit direction deviation distribution, adjacent interference distribution, edge clearance distribution, stacking risk distribution, adjacent component coupling distribution, and area ratio of components in each region based on continuous images. It then identifies a portion of the regions as areas that can be rearranged and improved, and another portion as areas that cannot be captured temporarily.

[0048] For the region corresponding to the rearranged liftable class, the control path decision unit generates a first rearrangement vibration sequence based on the main distribution direction of the deviation between the attitude convergence direction and the exit direction. This sequence includes a directional micro-vibration stage, a stop-vibration re-sampling stage, and a correction vibration stage. The duration of the directional micro-vibration stage is preferably set to 0.18 s to 0.32 s, with the vibration direction consistent with the attitude convergence direction, aiming to quickly bring the components closer to the allowable grasping posture set. The duration of the stop-vibration re-sampling stage is preferably set to 0.08 s to 0.12 s. In this stage, the flexible vibrating disk 5 stops vibrating, and the vision camera 6 resamples the region to confirm whether the corresponding target has been transformed from the rearranged liftable class to a stable graspable class. The duration of the vibration correction phase is preferably set to 0.12 s to 0.20 s. Its main function is to correct the deviation of the exit direction of the parts when they leave the disk and enter the grasping path, so that the target after the first rearrangement is easier for the robot 7 to grasp along the predetermined entry direction.

[0049] For areas corresponding to temporarily ungraspable zones, the control path decision unit generates a second rearrangement vibration sequence based on the stacking risk distribution, adjacent component coupling distribution, area proportion, and number of consecutive ineffective grasps. This sequence includes an enhanced decoupling vibration phase, a diffusion vibration phase, and a boundary evacuation vibration phase. The duration of the enhanced decoupling vibration phase is preferably set to 0.22 s to 0.35 s, primarily used to remove local overlaps and compressions; the duration of the diffusion vibration phase is preferably set to 0.18 s to 0.28 s, primarily used to increase the spacing between targets and reduce adjacent component coupling; the duration of the boundary evacuation vibration phase is preferably set to 0.10 s to 0.18 s, primarily used to disperse clustered targets near the disk boundary, reducing the probability of persistent dead zones forming at the edge.

[0050] During the execution of the second rearrangement vibration sequence, the control path decision unit restricts robot 7 from entering the grasping window of the corresponding area. After the sequence is completed, the disk state evolution unit re-determines the category. To verify the advantages of the present invention, four groups of test objects were set up. The invention example uses a phased first rearrangement vibration sequence and a second rearrangement vibration sequence generated based on regional distribution characteristics; Comparative Example 1 uses a uniform continuous vibration mode across the entire disk, without distinguishing between the areas corresponding to the rearrangement liftable class and the areas corresponding to the temporarily ungraspable area; Comparative Example 2 only uses directional micro-vibration and stop-vibration re-beat, without setting a correction vibration stage or a second rearrangement vibration sequence; Comparative Example 3 distinguishes between two types of areas, but the second rearrangement vibration sequence only uses a single enhanced vibration, excluding the diffusion vibration stage and the boundary evacuation vibration stage. Each group of tests runs continuously for 60 minutes, repeated 3 times, and the average value is taken as the result.

[0051] The specific results are as follows: Table 3: High-speed loading test results of components for electrically released door locks under different rearrangement vibration strategies Average percentage of stable crawlable classes after rearrangement / % 73.6 58.9 64.7 68.1 Rearrangement can improve the percentage of classes transitioning to stable classes / % 49.8 21.6 35.4 42.3 The rate of decrease in the area of ​​areas temporarily uncaptible is / %. 57.2 19.4 28.7 41.5 Edge clustering target reduction rate / % 61.8 18.9 24.6 36.2 Success rate of first effective crawl after rearrangement / % 97.9 90.8 93.6 95.1 Missed capture rate / % 0.4 2.7 1.6 1.0 Double capture rate / % 0.2 1.1 0.7 0.5 Clamping instability rate / % 0.3 1.5 0.9 0.8 Rearrangement beat recovery time / s 7.8 18.6 13.9 10.8 Incidence of repeated abnormal regions / % 1.9 12.7 7.4 4.8 Average feeding cycle time / piece per minute 53.4 46.2 49.1 51.0 Number of manual interventions per 60 consecutive minutes 0 4 2 1 As shown in Table 3, the invention significantly outperforms the three comparative examples in all key indicators, indicating that the first and second rearrangement vibration sequences are not simply refinements of vibration parameters, but rather core technical means for targeted reconstruction of the disk surface state. Firstly, considering the indicators of "average percentage of stable graspable classes after rearrangement" and "percentage of rearranged liftable classes transitioning to stable classes," the invention achieves 73.6% and 49.8% respectively, significantly higher than Comparative Example 1's 58.9% and 21.6%, and also higher than Comparative Examples 2 and 3. This demonstrates that for the rearranged liftable class region, relying solely on uniform continuous vibration or simply retaining simplified micro-vibration processes is insufficient to rapidly converge the target components to the allowed grasping posture set and allowed disk exit direction set; only by setting a stop-vibration re-beating stage after directional micro-vibration, and further correcting the disk exit direction deviation through vibration correction, can the grasping usability of the rearranged target be significantly improved. Although Comparative Example 2 retained directional micro-vibration and vibration-stopping re-shooting, the lack of a correction vibration stage resulted in some components, while their posture had improved, still exhibiting deviations in the off-disk path direction. Consequently, the initial effective grasping success rate after rearrangement was only 93.6%, lower than the 97.9% of the Invention Example. Secondly, regarding the "reduction rate of temporarily ungraspable area" and "reduction rate of edge-gathered targets," the Invention Example achieved 57.2% and 61.8% respectively, significantly higher than Comparative Example 1's 19.4% and 18.9%, and also superior to Comparative Example 3's 41.5% and 36.2%. This indicates that for temporarily ungraspable areas, simply using enhanced vibration is insufficient to continuously eliminate local overlap and edge congestion. While enhanced vibration can disperse some overlapping targets in a short time, it often pushes more components towards the disk boundary, causing edge gathering to reform. The Invention Example employs a three-stage combination of enhanced decoupling vibration, diffusion vibration, and boundary dispersal vibration, which removes local overlap, increases regional spacing, and further weakens boundary accumulation, thus more effectively reducing the area of ​​temporarily ungraspable areas. Furthermore, the combined data on the empty grasp rate, double grasp rate, gripping instability rate, and recurrence of abnormal areas show that the invention example has rates of 0.4%, 0.2%, 0.3%, and 1.9%, respectively, significantly lower than the respective ratios. This indicates that the difference between the two types of vibration sequences lies not only in the "different vibration intensity" but also in their different processing logics for abnormal areas on the disk surface. The first rearrangement vibration sequence focuses on areas that can be "transformed into graspable targets in a short time," aiming to improve the attitude and orientation quality of individual targets; the second rearrangement vibration sequence focuses on areas that "cannot be directly grasped in a short time and have severe local coupling," aiming to first relieve regional congestion and boundary dead zones. Because the invention example can adopt different rearrangement strategies for different areas, robot 7 faces not a set of chaotic targets in subsequent grasping cycles, but a set of targets after directional reconstruction, thus significantly reducing empty grasp, double grasp, and gripping instability problems.Finally, the "recovery time after rearrangement," "average feeding cycle time," and "number of manual interventions in 60 minutes" further demonstrate that the invention, while maintaining high grasping quality, did not sacrifice operational efficiency; instead, it achieved faster cycle time recovery and higher continuous operational stability. The invention's recovery time after rearrangement was only 7.8 seconds, with an average feeding cycle time of 53.4 pieces per minute and zero manual interventions in 60 minutes. In contrast, Comparative Example 1 had a recovery time of 18.6 seconds, an average feeding cycle time of only 46.2 pieces per minute, and required four manual interventions. This indicates that while uniform continuous vibration across the entire disk may seem to simplify the control process, it creates more unpredictable coupling states on the disk surface, causing the system to require a longer time to re-establish the distribution of graspable targets. As shown in Table 3, the technical solution generates a first and second rearrangement vibration sequence in a targeted manner based on the orientation convergence direction, the distribution of deviation in the exit direction, the distribution of adjacent interference, the distribution of edge clearance, the distribution of stacking risks, the distribution of adjacent coupling, and the area ratio of the region. This transforms the rearrangement action from a traditional unified vibration mode into a regionalized, phased, and targeted control process. As a result, it achieves simultaneous improvements in multiple dimensions, such as the ability to form stable graspable classes, the ability to clear abnormal areas, the grasping success rate, and the continuous operating cycle time. This effect cannot be naturally obtained by existing single vibration strategies, and it fully demonstrates the openness and technical advantages of the technical solution.

[0052] Example 5 like Figures 1-5 As shown, after completing the reconstruction of the single-variety operation constraint set, it enters the initialization calibration mode; In the initialization calibration mode, the parts to be loaded in the control bin 2 are conveyed to the flexible vibratory feeder 5 by the belt conveyor mechanism 3 in a preset initial batch. The flexible vibratory feeder 5 is controlled to execute a preset benchmark vibration sequence, and the initial stable graspable category ratio, the initial rearrangeable liftable category ratio, and the initial temporarily ungraspable area ratio are formed based on the initial feeder surface image to determine the initial control path corresponding to the current target model. The target screening threshold, the first rearrange vibration sequence, the second rearrange vibration sequence, the detection judgment threshold range, and the transfer cycle interval are initialized and corrected through at least one initial grasping cycle. In continuous feeding mode, the predicted amount of grippable parts is formed based on the number of stable grippable types, the expected conversion number of rearrangeable liftable types after executing the first rearrangement vibration sequence, the gripping cycle of robot 7, the transfer cycle of horizontal conveying mechanism 10, and the area ratio of temporarily ungripable areas. When the predicted remaining amount of grabbable parts is lower than the first preset threshold and the area of ​​the temporarily ungrabable area is lower than the second preset threshold, the material bin 2 is controlled to replenish material to the flexible vibrating plate 5 via the belt conveyor mechanism 3. When the predicted remaining amount of grabbable parts is lower than the first preset threshold and the area of ​​the temporarily ungrabable area is higher than or equal to the second preset threshold, the replenishment is delayed and the second rearrangement instruction is executed first. After the second rearrangement vibration sequence is completed, the predicted remaining amount of grabbable parts and the area ratio of temporarily ungrabable areas are recalculated, and replenishment is performed when the replenishment conditions are met. In this embodiment, to verify the tooling constraint reconstruction, initialization calibration, and predictive replenishment capabilities of the control system in a production change scenario, the "locking link A" in the automotive electric release door lock was selected as the original operating component, and the "locking link C" was selected as the target component after the production change for testing.

[0053] Locking link C is similar to locking link A in overall length and local hole layout, but the head offset direction, assembly hole orientation and stable clamping area of ​​locking link C are different from those of locking link A. Therefore, when changing production on the same feeding method, if only the mechanical replacement of the gripper and fixed tooling is performed without rebuilding the tooling constraints and initializing the running boundary, problems such as visual recognition passing but not being able to stably fall into the fixed tooling 12 after gripping, decreased first inspection pass rate of inspection camera 9, and disordered replenishment timing are very likely to occur.

[0054] The device includes a frame 1, a hopper 2, a belt conveyor mechanism 3, a drive motor 4, a flexible vibrating plate 5, a vision camera 6, a robot 7, a gripper 8, a detection camera 9, a horizontal handling mechanism 10, a conveyor belt 11, a fixed fixture 12, and a control unit.

[0055] When production changes begin, the continuous feeding process of the original locking link A is stopped first. The original gripper 8 and the original fixed fixture 12 are removed and replaced with new grippers 8 and new fixed fixture 12 that match the locking link C. The clamping opening range of the new gripper 8 is set to 15 mm to 20 mm, and the clamping action center is set 2.1 mm off the head side of the geometric center of the length direction of the locking link C. The allowable receiving direction of the new fixed fixture 12 is set so that the assembly hole faces the side where the inspection camera 9 is located, and the offset direction of the link head faces the side where the horizontal conveying mechanism 10 is located.

[0056] After the replacement is completed, the tooling constraint establishment unit immediately reads the tooling type information, allowed receiving direction, and limit positioning relationship of the new fixed tooling 12. Simultaneously, it reads the clamping opening state, clamping posture state, and clamping action center position of the new gripper 8, constructing a receiving constraint domain and a clamping constraint domain. The two are then intersected and mapped to generate a joint feasible domain for the locking link C. Subsequently, based on this joint feasible domain, the tooling constraint establishment unit performs consistency screening on the disk surface posture category, discharge direction category, and receiving direction category of the subsequent parts to be loaded onto the flexible vibratory feeder 5. Only the category information that simultaneously falls within the joint feasible domain is written into the single-product operation constraint set, forming a set of allowed gripping postures, a set of allowed discharge directions, and a set of allowed receiving directions. The detection judgment threshold range and transfer cycle interval are updated simultaneously. To prevent the system from relying on repeated manual trial gripping after production changeover, this implementation immediately enters the initialization calibration mode after completing the reconstruction of the single-product operation constraint set. In this mode, the surplus prediction and replenishment unit first controls the drive motor 4 to drive the belt conveyor mechanism 3 to transport the locking linkage C in the hopper 2 to the flexible vibratory feeder 5 according to the preset initial batch. In this embodiment, the preset initial batch is set to 22 pieces. Subsequently, the control path decision unit controls the flexible vibratory feeder 5 to execute the preset benchmark vibration sequence. The benchmark vibration sequence consists of a 0.9s basic dispersion vibration stage, a 0.2s stop vibration and re-shoot stage, and a 0.4s attitude adjustment vibration stage. The feeder state evolution unit performs initial category determination on the parts on the flexible vibratory feeder 5 based on the initial feeder surface images continuously acquired by the vision camera 6, and generates the initial stable graspable category ratio, the initial rearrangeable liftable category ratio, and the initial temporarily ungraspable area ratio. The control path decision unit determines the initial control path based on the above three initial ratios and drives the robot 7 to execute the initial grasping cycle. The grasping result linkage write-back unit statistically analyzes empty grasping, off-center grasping, double grasping, and unstable clamping during the initial grasping cycle, and performs initial corrections on the target screening threshold, the first rearrangement vibration sequence, the second rearrangement vibration sequence, the detection and judgment threshold range, and the transfer cycle interval. After completing the above corrections, the remaining quantity prediction and replenishment unit starts working. It does not replenish based solely on the number of visible parts on the flexible vibrating plate 5, but rather on the current number of stable graspable parts, the expected conversion number of rearrangeable liftable parts after executing the first rearrangement vibration sequence, the grasping cycle time of the robot 7, the transfer cycle time of the horizontal transport mechanism 10, and the area ratio of the temporarily ungraspable area, to generate the predicted remaining quantity of graspable parts for the next replenishment cycle.

[0057] In this embodiment, the prediction of the remaining amount of graspable parts adopts the following calculation logic: the current stable graspable class quantity is denoted as Ns, the expected conversion quantity of rearrangeable class is denoted as Nr, the expected conversion coefficient is denoted as η, the grasping cycle of robot 7 is denoted as Vg, the transfer cycle of horizontal conveying mechanism 10 is denoted as Vt, and the replenishment prediction time window is denoted as Tf. Then, the predicted remaining amount of graspable parts Qp is estimated according to "Qp = Ns + η × Nr - min(Vg, Vt) × Tf". Wherein, η is dynamically updated according to the actual transformation of the first rearranged vibration sequence in the previous time window. In this embodiment, η is initially set to 0.62.

[0058] If Qp is lower than the first preset threshold and the proportion of the temporarily ungrabable area is lower than the second preset threshold, the remaining quantity prediction and replenishment unit immediately controls the belt conveyor 3 to replenish materials according to the preset batch. If Qp is lower than the first preset threshold but the proportion of the temporarily ungrabable area is higher than or equal to the second preset threshold, the replenishment is delayed and the control path decision unit is called first to output the second rearrangement instruction. The second rearrangement vibration sequence is executed first for the corresponding area, and the replenishment is executed after the proportion of the temporarily ungrabable area decreases.

[0059] To verify the effectiveness of the present invention, four groups of test subjects were set up: The invention employs a complete control strategy that combines joint feasible domain establishment, initial calibration mode, and predictive feeding. Comparative Example 1 adopts the method of directly calling fixed formula parameters after changing tooling and triggering material replenishment with low material level threshold, without establishing a joint feasible domain and without entering the initialization calibration mode; Comparative Example 2 uses a joint feasible domain to establish the model, but does not enter the initialization calibration mode. After completing the parameter update, it directly enters continuous feeding. Comparative Example 3 adopts a joint feasible domain establishment and initial calibration mode, but the material replenishment is still based only on the number of currently visible parts, without considering the expected conversion quantity and the proportion of the area of ​​the temporarily ungraspable region.

[0060] Each experiment involved three consecutive production changes, with each change followed by a 60-minute continuous run. The average value was recorded in the table below: Table 4: Comparison of Control Mechanism Test Results under Production Change Conditions Joint feasible domain establishment time / s 11.8 0 10.9 11.4 Initial calibration completion time / s 142 0 0 151 Initial gripping cycle / time required to achieve continuous feeding mode 9 0 0 10 Stable operating time after production change / s 186 524 368 247 Effective capture rate in the first 10 minutes after the production change / % 97.6 88.9 93.1 95.2 First inspection pass rate (%) within the first 10 minutes after production change 98.8 93.7 96.2 97.1 Mis-transfer rate of mixed materials in the first 10 minutes after production change / % 0.05 0.61 0.22 0.14 Predicted replenishment accuracy rate / % 96.4 71.8 74.3 82.7 Ineffective feeding times / time 0 5 4 2 Average area percentage of areas temporarily unscrapable / % 8.7 22.9 15.6 11.8 Average continuous feeding cycle time / piece per minute 53.1 45.8 49.4 51.0 Beat fluctuation coefficient / % 3.1 10.7 6.9 5.2 Number of manual interventions per 60 minutes after delivery 0 4 2 1 As shown in Table 4, the invention significantly outperforms the three comparative examples in terms of stable operating speed after production change, grabbing quality in the first 10 minutes, material replenishment rationality, continuous feeding cycle time, and manual intervention control. This indicates that its corresponding technical solution is not a simple parameter modification of the existing automatic feeding line, but rather a complete machine control mechanism strongly coupled with the production change scenario constructed through "tooling constraint reconstruction, initial calibration, predictive replenishment, and continuous mode switching." First, considering the four indicators—joint feasible domain establishment time, initial calibration completion time, initial grabbing cycle required to reach continuous feeding mode, and stable operating time after production change—although the invention adds an 11.8 s joint feasible domain establishment process and a 142 s initial calibration process, the stable operating time after production change is only 186 s, significantly better than Comparative Example 1's 524 s and Comparative Example 2's 368 s, and also better than Comparative Example 3's 247 s. This indicates that the newly added constraint reconstruction and initial calibration will not slow down the production changeover efficiency. On the contrary, it can avoid a large number of subsequent invalid trial grabs, parameter rollbacks and manual interventions, thus shortening the total time from the start of the production changeover to entering a stable and continuous feeding phase.

[0061] Comparative Example 1, by only calling fixed formula parameters and not establishing a joint feasible region based on the physical constraints of the new fixed tooling 12 and the new gripper 8, resulted in the system continuing to use incompatible gripping, receiving, and detection boundaries for a considerable period after the product change, thus achieving the longest stable operating time. Comparative Example 2, although establishing a joint feasible region, lacked prior correction for the initial disk surface state and initial ineffective gripping distribution of the new target model due to the absence of initial calibration. Therefore, it still needed to rely on actual anomalies to gradually correct these issues during the initial phase of continuous operation, resulting in a significantly longer stable operating time compared to the Invention Example.

[0062] Secondly, considering the effective gripping rate, first-pass rate, and mixed material mistransfer rate in the first 10 minutes after the production change, the invention achieved 97.6%, 98.8%, and 0.05% respectively, significantly better than the other three groups. This data shows that the joint feasible domain established by the tooling constraint establishment unit is not an abstract software screening condition, but truly transforms the allowable receiving direction of the fixed tooling 12 and the clamping boundary of the gripper 8 into the preconditions for gripping legality, thus enabling the system to operate according to the physical boundaries of the new target model within the first 10 minutes after the production change. Meanwhile, the initial calibration mode further exposes and corrects the most likely problems that occur in the early stages of production changeover by using initial material replenishment, reference vibration, initial category determination, and initial gripping cycle. Therefore, the invention has a particularly obvious advantage in terms of first-inspection pass rate and mixed material mistransfer rate.

[0063] Furthermore, in terms of the three indicators of prediction accuracy, number of invalid replenishments, and average area ratio of temporarily ungraspable areas, the invention achieved 96.4%, 0 times, and 8.7% respectively, which is significantly better than Comparative Example 1's 71.8%, 5 times, and 22.9%, and also better than Comparative Example 2 and Comparative Example 3.

[0064] This fully demonstrates that replenishment based on remaining quantity prediction is not simply a change in the calculation method of "replenishing when the remaining quantity is low." Instead, it incorporates the number of stable graspable categories, the expected conversion quantity of rearrangeable and upgradable categories, the robot's grasping cycle time (7), the horizontal transport mechanism's transfer cycle time (10), and the proportion of temporarily ungraspable areas into the replenishment decision. This ensures that replenishment behavior truly matches the availability of the pallet and subsequent processing capacity. Although Comparative Example 3 also entered the initial calibration mode, replenishment was still performed based on the currently visible quantity, without considering the expected conversion quantity and the proportion of temporarily ungraspable areas. Therefore, its number of invalid replenishments was still 2, indicating that if replenishment continues before the temporarily ungraspable areas on the pallet are cleared, local congestion and boundary dead zones will be further amplified.

[0065] Finally, considering the average continuous feeding cycle time, cycle time fluctuation coefficient, and number of manual interventions in the 60 minutes after production changeover, the invention example achieved 53.1 pieces per minute, 3.1%, and 0 interventions, respectively, which are superior to Comparative Example 1's 45.8 pieces per minute, 10.7%, and 4 interventions, and also superior to the other comparative examples. This indicates that the present invention, through combined feasible domain generation, initial calibration switching, and predictive feeding control, enables the entire machine to not only enter a usable state more quickly after production changeover but also maintain a more stable continuous operating cycle time. In particular, the cycle time fluctuation coefficient decreased from 10.7% in Comparative Example 1 to 3.1% in the invention example, indicating that the invention example does not only perform better in individual gripping cycles but also forms a more stable operating boundary in the sense of continuous operation.

[0066] Based on the results in Table 4, it can be concluded that organizing "constraint reconstruction after tooling and gripper replacement", "boundary convergence in initial calibration mode" and "predictive material replenishment synchronized with the plate status and cycle time" into a unified closed loop not only significantly shortens the time to reach stable operation after production changeover, but also simultaneously improves the effective gripping rate, first inspection pass rate, material replenishment accuracy and continuous cycle time stability in the early stage of production changeover.

[0067] A control system for a high-speed feeding method for components of an electrically released door lock, wherein the control system is housed in a control unit and is connected to a drive motor 4, a flexible vibratory feeder 5, a vision camera 6, a robot 7, a detection camera 9, and a horizontal conveying mechanism 10, characterized in that the control system includes: The tooling constraint establishment unit is used to determine the target model of the part to be loaded based on the tooling type information of the fixed tooling 12, the allowed receiving direction, and the clamping opening state and clamping posture state of the gripper 8, and to establish a single-product operation constraint set corresponding to the target model. The disk state evolution unit is used to construct a disk state sequence of the parts to be loaded on the flexible vibrating disk 5 based on the disk images continuously acquired by the vision camera 6, and to generate stable graspable class, rearrangeable liftable class and temporarily ungraspable area. The control path decision unit is used to output direct grabbing instructions, first rearrangement instructions, or second rearrangement instructions based on the single-variety operation constraint set and the tray classification results, and control the robot 7 and the flexible vibrating plate 5 to perform the corresponding operations. The gripping result linkage write-back unit is used to determine the validity of the gripping result based on the execution status of the gripper 8, the detection result of the detection camera 9, and the transfer response result of the horizontal conveying mechanism 10. When the gripping result is determined to be invalid, the front-end material handling parameters and the back-end detection or transfer parameters are simultaneously corrected. The remaining quantity prediction and replenishment unit is used to predict the remaining quantity of grippable parts on the flexible vibratory feeder 5 based on the number of stable grippable types, the gripping cycle of the robot 7 and the transfer cycle of the horizontal conveying mechanism 10. When the remaining quantity of grippable parts is lower than a preset threshold, the unit controls the drive motor 4 to drive the belt conveyor mechanism 3 to transport the parts to be loaded in the hopper 2 to the flexible vibratory feeder 5.

[0068] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0069] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the present invention. Various changes and modifications can be made to the present invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed.

Claims

1. A high-speed feeding method for components of an electrically released door lock, applied to an electrically released door lock feeding device, the electrically released door lock feeding device comprising a flexible vibratory feeder (5), a vision camera (6), a robot (7), a gripper (8), a fixed fixture (12), a detection camera (9), a horizontal conveying mechanism (10), a hopper (2), and a belt conveyor mechanism (3), characterized in that, include: After detecting that the fixed fixture (12) or the gripper (8) has been replaced, the fixture type information, allowable bearing direction and limit positioning relationship of the fixed fixture (12), as well as the clamping opening state, clamping posture state and clamping action center position of the gripper (8) are read. The bearing constraint domain and the clamping constraint domain are constructed respectively, and the intersection mapping of the bearing constraint domain and the clamping constraint domain is performed to obtain the joint feasible domain corresponding to the current target model. Based on the joint feasible domain, a single product operation constraint set is established. Based on at least three frames of disk surface images continuously acquired by the vision camera (6), the trajectory association of the parts to be loaded on the flexible vibrating disk (5) is performed to determine the pose change trend, the discharge direction change trend and the adjacent interference state of each part to be loaded. Combined with the single-product operation constraint set, each part to be loaded or disk surface area is classified and judged to form a stable graspable class, a rearrangeable liftable class and a temporarily ungraspable area. According to the classification and judgment results, a unique corresponding control path is output. For the stable graspable class, a direct grasping command is output. For the rearrangeable liftable class, a first rearrangement command is output. For the temporarily ungraspable area, a second rearrangement command is output. The robot (7) is restricted from entering the grasping window of the corresponding area. After the robot (7) completes the grasping and transfer, the validity of this grasping is determined based on the opening and closing displacement change of the gripper (8), the gripping drive response change, the detection results of the detection camera (9) on the parts located on the fixed fixture (12), and the transfer response results of the horizontal transport mechanism (10). When a non-valid grab is determined, the front-end material grabbing parameters and the back-end detection or transfer parameters are adjusted simultaneously, and the corresponding tray area is remarked as a rearrangeable liftable area or a temporarily ungrabable area. After completing the reconstruction of the single-product operation constraint set, it enters the initialization calibration mode, and after meeting the switching conditions corresponding to the effective grasping ratio and the area ratio of the temporarily ungraspable area, it enters the continuous feeding mode. In continuous feeding mode, the material bin (2) is fed to the flexible vibrating plate (5) via the belt conveyor mechanism (3) according to the predicted remaining amount of grabbable parts and the area ratio of the temporarily ungrabable area, or the feeding is delayed and the second rearrangement instruction is executed first.

2. The high-speed feeding method for components of an electrically released door lock according to claim 1, characterized in that, The establishment of a single-product operational constraint set corresponding to the current target model based on the joint feasible domain includes: The intersection of the receiving constraint domain and the clamping constraint domain is mapped to form a joint feasible domain corresponding to the current target model. Based on the joint feasible domain, the consistency screening is performed on the disk surface posture category, disk discharge direction category and receiving direction category of the parts to be loaded on the flexible vibratory disk (5); Only the attitude category, exit direction category, and acceptance direction category that simultaneously fall into the joint feasible domain are written into the single-product operation constraint set to form the allowed grab attitude set, allowed exit direction set, and allowed acceptance direction set; And after the fixed fixture (12) or the gripper (8) is replaced, the detection judgment threshold range and the transfer cycle range are updated synchronously.

3. A high-speed feeding method for components of an electrically released door lock according to claim 2, characterized in that, The classification and determination of the parts to be loaded or the disk area based on at least three disk surface images includes: Based on at least three frames of disk surface images continuously acquired by the vision camera (6), the trajectory association of each component to be loaded on the flexible vibrating disk (5) is performed. Combined with the gripping opening range of the gripper (8), the grasping entry direction of the robot (7), the allowable receiving direction of the fixed fixture (12), and the single-product operation constraint set, posture constraint matching value, disk exit direction constraint matching value, receiving direction constraint matching value, and adjacent interference risk value are established for each component to be loaded. When the attitude constraint matching value of a certain component to be loaded is higher than the first threshold, the discharge direction constraint matching value is higher than the second threshold, the receiving direction constraint matching value is higher than the third threshold, and the adjacent interference risk value is lower than the fourth threshold, it is determined to be a stable and graspable type. When a certain component to be loaded does not meet the determination criteria for a stable graspable class, but its posture change trend and position migration trend in adjacent sampling periods converge toward the set of allowed grasping postures, the set of allowed ejection directions, and the set of allowed receiving directions, it is determined to be a rearrangeable liftable class. If the parts to be loaded in a certain area do not meet the criteria for stable graspability in multiple consecutive sampling periods, and the stacking risk and adjacent coupling of the parts to be loaded in that area are continuously higher than the preset threshold, then the area is determined to be a temporarily ungrabable area.

4. A high-speed feeding method for components of an electrically released door lock according to claim 3, characterized in that, The first rearrangement instruction is used to control the flexible vibratory plate (5) to vibrate according to the first rearrangement vibration sequence. The first rearrangement vibration sequence is generated based on the posture convergence direction, discharge direction deviation distribution, adjacent interference distribution and edge clearance distribution of the parts to be loaded in the corresponding area of ​​the rearrangeable lifting class. The first rearranged vibration sequence includes a directional micro-vibration stage along the attitude convergence direction, a stop-vibration and re-shooting stage spaced apart from the directional micro-vibration stage, and a correction vibration stage for correcting the deviation of the disk exit direction. In the stop-vibration and re-shooting stage, the vision camera (6) is called to resample the corresponding area, and the first rearranged vibration sequence is maintained or switched to a direct grasping command based on the resampling result. The second rearrangement instruction is used to control the flexible vibratory plate (5) to vibrate according to the second rearrangement vibration sequence. The second rearrangement vibration sequence is generated based on the stacking risk distribution, adjacent coupling distribution, area ratio and number of consecutive ineffective grasps of the parts to be loaded in the corresponding area of ​​the temporarily ungrabable area. The second rearranged vibration sequence includes an enhanced decoupling vibration stage, a diffusion vibration stage, and a boundary evacuation vibration stage; During the second rearrangement vibration sequence, the robot (7) is restricted from entering the grasping window of the corresponding area, and the corresponding area is reclassified after the second rearrangement vibration sequence is completed.

5. A high-speed feeding method for components of an electrically released door lock according to claim 4, characterized in that, The determination of the validity of this crawl includes: After the robot (7) completes a gripping and transfer, based on the opening and closing displacement change of the gripper (8), the gripping drive response change, the detection result of the detection camera (9) on the parts located on the fixed fixture (12), and the transfer response time of the horizontal transport mechanism (10), the gripping result is determined as empty gripping, biased gripping, double gripping, unstable gripping, or effective gripping. When the judgment result is empty gripping, biased gripping, double gripping or unstable clamping, at least one of the target screening threshold of the front-end material picking side, the first rearrangement vibration sequence, the second rearrangement vibration sequence and the gripping path of the robot (7) are adjusted synchronously, as well as at least one of the detection judgment threshold range and the transfer rhythm range of the back-end detection or transfer side, and the corresponding plate area where ineffective gripping occurs is remarked as a rearrangeable liftable area or a temporarily ungripable area; When an invalid grasp occurs in the same area within a preset number of consecutive times, the robot (7) is restricted from entering the grasping window of that area, and the control priority of executing the second rearrangement instruction in that area is increased.

6. A high-speed feeding method for components of an electrically released door lock according to claim 5, characterized in that, After completing the reconstruction of the single-variety operating constraint set, the initialization calibration mode is entered. In the initialization calibration mode, the parts to be loaded in the hopper (2) are controlled to be transported to the flexible vibrating plate (5) in a preset initial batch via the belt conveyor mechanism (3). The flexible vibrating plate (5) is controlled to execute a preset reference vibration sequence. Based on the initial plate surface image, the initial stable graspable category ratio, the initial rearrangeable liftable category ratio, and the initial temporarily ungraspable area ratio are formed to determine the initial control path corresponding to the current target model. The target screening threshold, the first rearrange vibration sequence, the second rearrange vibration sequence, the detection judgment threshold interval, and the transfer beat interval are initialized and corrected through at least one initial grasping cycle. In the continuous feeding mode, the predicted amount of grippable parts is formed based on the number of stable grippable types, the expected conversion number of rearrangeable liftable types after executing the first rearrangement vibration sequence, the gripping rhythm of the robot (7), the transfer rhythm of the horizontal transport mechanism (10), and the area ratio of the temporarily ungripable area. When the predicted amount of grabbable parts is lower than the first preset threshold and the area ratio of the temporarily ungrabable area is lower than the second preset threshold, the hopper (2) is controlled to replenish the flexible vibrating plate (5) via the belt conveyor mechanism (3); When the predicted remaining amount of grabbable parts is lower than the first preset threshold and the area ratio of the temporarily ungrabable area is higher than or equal to the second preset threshold, the replenishment is delayed and the second rearrangement instruction is executed first. After the second rearrangement vibration sequence is completed, the predicted remaining amount of grabbable parts and the area ratio of the temporarily ungrabable area are recalculated, and replenishment is performed when the replenishment conditions are met.

Images