Flexible sealing ring assembling equipment and method based on mechanical vision technology

Abstract

The invention relates to the field of workpiece assembling, in particular to sealing ring flexible assembling equipment and method based on the mechanical vision technology, and the equipment comprises a vibration disc, a first deep learning shooting module, a second deep learning shooting module, a four-axis robot, a transfer platform and an assembling jig. The vibration disc is used for containing a to-be-assembled sealing ring. The first deep learning shooting module is arranged on the top side of the vibration disc and used for recognizing first coordinate data of the sealing ring and sending the first coordinate data to the four-axis robot. The four-axis robot is used for picking up the sealing ring from the vibration disc and conveying the sealing ring to the transfer platform for secondary positioning according to the first coordinate data, and after secondary positioning is completed, the sealing ring is conveyed to a base of the assembling jig to be assembled; and the second deep learning shooting module is arranged on the top side of the vibration disc and is used for carrying out flying shooting detection on an assembled finished product in the assembly jig according to a preset detection algorithm. The equipment overcomes the problem of low assembly quality of the electronic cigarette in the prior art.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of workpiece assembly. More particularly, the present application relates to a flexible sealing ring assembly device and method based on mechanical vision technology. BACKGROUND

[0002] The sealing ring of the electronic cigarette is a key part mainly used for preventing leakage of tobacco tar and protecting internal components from external environment, which is usually arranged at the connection position of the cartridge and the cigarette rod, the oil tank sealing position or the atomizer sealing position.

[0003] For the assembly of the electronic sealing ring, the Chinese patent application CN223084141U discloses a sealing ring assembly device, which includes a rack, a ring supporting device and a feeding device. The ring supporting device and the feeding device are both mounted on the upper end surface of the rack. The ring supporting device includes a first driving assembly and multiple ring supporting mechanisms, and the multiple ring supporting mechanisms are configured to be expanded or contracted under the driving of the first driving assembly. The feeding device is mounted on the upper end surface of the rack and includes a material taking mechanism and a conveying mechanism. The material taking mechanism is provided with multiple material taking structures corresponding to the multiple ring supporting mechanisms, and the conveying mechanism is used for accommodating workpieces. When the ring supporting mechanism is expanded, the sealing ring placed on the ring supporting mechanism is expanded, and the multiple material taking structures can take the expanded sealing ring and assemble it with the corresponding workpiece. The sealing ring assembly device reduces processing cost and improves processing efficiency.

[0004] Although the above sealing ring assembly device can automatically assemble the sealing ring of the electronic cigarette, it belongs to an open-loop assembly process based on action teaching during the sealing ring assembly process, only fixed teaching actions are performed, and it cannot identify the assembly defects of the sealing ring (including defects before assembly and defects after assembly) and correct or exclude the assembly defects, which makes the quality of the finally assembled electronic cigarette low. SUMMARY

[0005] To solve the above technical problem of low quality of electronic cigarette assembly, the present application discloses a flexible sealing ring assembly device and method based on mechanical vision technology.

[0006] In a first aspect, the present application discloses a flexible sealing ring assembly device based on mechanical vision technology, which comprises a vibrating disc, a first deep learning shooting module, a second deep learning shooting module, a four-axis robot, a transfer platform and an assembly jig; wherein, The vibrating disc is used to accommodate the sealing ring to be assembled; The first deep learning shooting module is arranged on the top side of the vibrating disc and is used to identify the first coordinate data of the sealing ring and send the first coordinate data to the four-axis robot; The four-axis robot is used for picking up the sealing ring from the vibration disc according to the first coordinate data and conveying to the transfer platform for secondary positioning, and after the secondary positioning is completed, the sealing ring is conveyed to the base of the assembly jig for assembly. The second deep learning shooting module is arranged on the top side of the assembly jig and is used for flying shot detection on the assembled products in the assembly jig according to a preset detection algorithm, and if the assembled product is detected as a defective product, a rejection instruction is sent to the master control device of the next work station.

[0007] Beneficial effects: The vibration disc of the equipment is used for accommodating the sealing ring, and the multiple sealing rings are laid out by vibration, avoiding that the four-axis robot picks up multiple sealing rings overlapped vertically at a time, causing assembly defects. The first deep learning shooting module accurately identifies the first coordinate data of the sealing ring by using the built-in deep learning algorithm, and sends the first coordinate data as feedback to the four-axis robot, so that the four-axis robot can accurately pick up the sealing ring. The transfer platform is used for assisting the sealing ring to be positioned secondly, and has the function of correcting the sealing ring. The four-axis robot is used for picking up the sealing ring and completing the sealing ring assembly process. The second deep learning shooting module is used for flying shot detection on the assembled products, so as to quickly promote the sorting of defective products. Compared with the prior art, the equipment avoids the defects of the electronic cigarette sealing ring in the assembly process in the ways of vibration laying, pre-assembly recognition and positioning, and post-assembly detection, and issues a screening instruction for the defective products, overcoming the technical problem of low quality of the existing electronic cigarette assembly.

[0008] Preferably, the first deep learning shooting module comprises a first deep learning camera and a first light source, the first light source is arranged on the side of the first deep learning camera, and the lens of the first deep learning camera and the first light source are both vertically directed towards the vibration disc.

[0009] Beneficial effects: The first deep learning camera is used for acquiring a two-dimensional laying image of the sealing ring, and the first light source is used for providing illumination for the first deep learning camera, so that the imaging of the first deep learning camera is clearer.

[0010] Preferably, the second deep learning shooting module comprises a second deep learning camera and a second light source, and the lens of the second deep learning camera and the second light source are both inclined towards the assembly jig.

[0011] Beneficial effects: The second deep learning camera is inclined towards the assembly jig to acquire an edge image of the assembled product (a sealing ring edge image), and the second light source is used for providing illumination for the second deep learning camera, so that the imaging of the second deep learning camera is clearer.

[0012] Preferably, a magnetic levitation guide rail is further included, and the assembly jig is arranged on the top of the magnetic levitation guide rail.

[0013] Preferably, the vibration disc adopts a flexible vibration disc.

[0014] In a second aspect, the present application further discloses a flexible sealing ring assembly method based on mechanical vision technology, which is used for the flexible sealing ring assembly device based on mechanical vision technology in the first aspect, and the method comprises the following steps: sending an identification instruction to the first deep learning shooting module to obtain first coordinate data of the sealing ring; predicting a first pick-and-place trajectory of the four-axis robot according to the first coordinate data and a current position of the four-axis robot, and sending an execution instruction of the first pick-and-place trajectory to the four-axis robot; sending a jig assembly instruction to the four-axis robot in response to completion of the execution instruction; sending a snapshot instruction to the second deep learning shooting module in response to completion of the jig assembly instruction to obtain a snapshot image; comparing the snapshot image with a standard template image according to a preset deep learning algorithm to calculate a matching degree of the snapshot image and the standard template image; determining that the assembly product is a defective product if the matching degree is lower than a threshold.

[0015] Beneficial effects: the method sends an identification instruction to the first deep learning shooting module first, so that the first deep learning shooting module takes a photo of the sealing ring to obtain first coordinate data of the sealing ring; then the first pick-and-place trajectory of the four-axis robot is predicted according to the first coordinate data and a current position of the four-axis robot to realize trajectory optimization of the four-axis robot and avoid damage or loss of the sealing ring caused by deviation or trajectory interference; after completion of the jig assembly instruction, the snapshot instruction is sent to the second deep learning shooting module and the deep learning algorithm is used to calculate the matching degree of the snapshot image and the standard template image, and the matching degree is used to determine whether the assembly product is a defective product, so that the defective product can be quickly screened out. Compared with the prior art, the method can realize high-quality and high-efficiency electronic cigarette sealing ring assembly.

[0016] Preferably, the deep learning algorithm adopts a convolutional neural network model or an adversarial network model.

[0017] Preferably, the comparison of the snapshot image and the standard template image comprises the following steps: collecting edge features of the snapshot image to obtain a first edge feature matrix; collecting edge features of the standard template image to obtain a second edge feature matrix; calculating a cosine similarity of the first edge feature matrix and the second edge feature matrix, and taking the cosine similarity as the matching degree.

[0018] Preferably, the jig assembly instruction comprises a cartridge-tobacco rod-sealing assembly instruction, an oil tank sealing assembly instruction and / or an atomizer sealing assembly instruction.

[0019] Beneficial effects: the method of the present application supports the seal assembly of the cartridge-tobacco rod-seal assembly, the oil tank seal assembly and the atomizer seal assembly, and is almost suitable for the seal assembly of all electronic cigarette parts.

[0020] Preferably, before issuing the identification instruction to the first deep learning shooting module, the method further comprises: issuing a vibration instruction to the vibrating disc for accommodating the sealing ring, so that the plurality of sealing rings are horizontally laid on the vibrating disc.

[0021] The beneficial effects of the present application are: (1) Compared with the prior art, the device of the present application avoids defects that may occur in the assembly process of the electronic cigarette sealing ring to the greatest extent through vibration paving, pre-assembly recognition positioning and post-assembly detection, and issues screening instructions for defective products, overcoming the technical problem of low quality of existing electronic cigarette assembly.

[0022] (2) Compared with the prior art, the method of the present application can realize high-quality and high-efficiency electronic cigarette sealing ring assembly. BRIEF DESCRIPTION OF DRAWINGS

[0023] The above and other objects, features and advantages of the exemplary embodiments of the present application will be more apparent from the following detailed description read in conjunction with the accompanying drawings, in which several embodiments of the present application are shown by way of example, and wherein the same or corresponding elements are referred to by the same or corresponding reference numerals, in which: Figure 1 is a structural schematic diagram of a sealing ring flexible assembly device based on mechanical vision technology in embodiment one of the present application; Figure 2 is an assembly schematic diagram of the sealing ring and the base of the assembly jig in embodiment one of the present application; Figure 3 is a flowchart of a sealing ring flexible assembly method based on mechanical vision technology in embodiment two of the present application.

[0024] BRIEF DESCRIPTION OF DRAWINGS: 100, machine table; 1, vibrating disc; 2, first deep learning shooting module; 3, second deep learning shooting module; 31, second deep learning camera; 32, second light source; 4, four-axis robot; 41, first rotating shaft; 42, second rotating shaft; 43, lifting shaft; 44, third rotating shaft; 45, pick-up head; 5, transfer platform; 6, assembly jig; 7, magnetic levitation guide rail. DETAILED DESCRIPTION

[0025] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. 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.

[0026] This embodiment discloses a flexible assembly device and method for sealing rings based on machine vision technology, which is used to solve the technical problem of low assembly quality of electronic cigarettes in the prior art.

[0027] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0028] Example 1 like Figure 1 As shown, this embodiment discloses a flexible assembly device for sealing rings based on machine vision technology, including a machine base 100, a vibratory plate 1, a first deep learning imaging module 2, a second deep learning imaging module 3, a four-axis robot 4, a transfer platform 5, an assembly fixture 6, and a magnetic levitation guide rail 7.

[0029] The machine 100 is used to support and fix the aforementioned vibratory plate 1, first deep learning imaging module 2, second deep learning imaging module 3, four-axis robot 4, transfer platform 5, assembly fixture 6, and magnetic levitation guide rail 7.

[0030] Furthermore, the vibratory feeder 1 is a flexible vibratory feeder, which can effectively absorb impacts when collisions or compression occur during the conveying / vibration of the sealing ring (generally a rubber ring), greatly reducing damage and scratches to the sealing ring and thus protecting the surface of the sealing ring. In this embodiment, two vibratory feeders 1 are arranged in a mirror-symmetrical manner, both used to accommodate the sealing rings, and can spread out the stacked sealing rings through vibration, avoiding the four-axis robot 4 from picking up multiple overlapping sealing rings at once, which would lead to assembly defects (i.e., assembling multiple sealing rings into one finished product).

[0031] Further, the first deep learning shooting module 2 of the embodiment is also mirror-symmetrically arranged in two, and is located at the top side of the corresponding vibration disc 1, and is used for identifying the first coordinate data of the sealing ring and sending the first coordinate data to the four-axis robot 4. More specifically, the first deep learning shooting module 2 comprises a first deep learning camera and a first light source (not shown in the figure), the first light source is arranged at the side of the first deep learning camera, and the lens of the first deep learning camera and the first light source are both vertically directed to the vibration disc 1. Through the above structure design, the first deep learning camera is used to obtain the two-dimensional paving image of the sealing ring, and based on the built-in deep learning algorithm, the first coordinate data (two-dimensional coordinates of the center) of the sealing ring is calculated. As for the first light source, it is used to provide light for the first deep learning camera, so that the imaging of the first deep learning camera is clearer.

[0032] Further, the four-axis robot 4 is provided with a first rotating shaft 41, a second rotating shaft 42, a lifting shaft 43, a third rotating shaft 44 and a pickup head 45 from the robot base to the pickup end. The pickup head 45 can adopt a flexible clamp. The four-axis robot 4 will pick up the sealing ring located in the vibration disc 1 to the transfer platform 5 for secondary positioning according to the first coordinate data. After the sealing ring completes the secondary positioning, as shown in the figure, the sealing ring is stretched by the flexible clamp, and the stretched sealing ring is transported to the periphery of the base of the assembly jig 6. When the flexible clamp is contracted, the sealing ring will slide downward based on the action of gravity and contraction force, and be assembled with the base (groove) of the assembly jig 6. Figure 2

[0033] It should be noted that in the embodiment, a plurality of ring grooves (not shown in the figure) are arranged on the transfer platform 5, which are adapted to the shape of the sealing ring. The middle part of the ring groove is hollow to support the above-mentioned pickup head 45. When the four-axis robot 4 picks up the sealing ring located in the vibration disc 1 to the transfer platform 5, the pickup head 45 is loosened, and if the position deviation of the sealing ring and the ring groove is not large, the sealing ring will automatically slide into the ring groove to realize secondary positioning.

[0034] Preferably, a vibrator can be arranged on the transfer platform 5, so that the sealing ring falls into the ring groove through vibration.

[0035] Preferably, a flexible scraper can also be arranged on the transfer platform 5, so that the sealing ring falls into the ring groove through the flexible scraper.

[0036] ​In the embodiment, the assembly jig 6 can be used as a carrier for the cartridge-tobacco rod components, oil tank assembly or atomizer, and can carry multiple components by providing multiple hole positions. The assembly jig 6 is arranged on the top of the magnetic levitation guide rail 7. Through the above structure design, the device of the present application realizes the non-damping conveying of the feeding and discharging by the magnetic levitation technology, which can avoid the violent vibration of the raw materials and finished products during the conveying process, thereby avoiding the deviation of the sealing ring assembly position. Compared with the prior art, the device of the present application improves the quality of electronic cigarette assembly by using the characteristics of magnetic levitation technology.

[0037] Further, the second deep learning shooting module 3 is arranged on the top side of the assembly jig 6, and is used for flying shot detection of the assembled finished product in the assembly jig 6 according to a preset detection algorithm. If the assembled finished product is detected as a defective product, a rejection instruction is sent to the master control device of the next work station. More specifically, the second deep learning shooting module 3 includes a second deep learning camera 31 and a second light source 32. The lens of the second deep learning camera 31 and the second light source 32 are both inclined towards the assembly jig 6.

[0038] Through the above structure design, the second deep learning camera 31 is inclined towards the assembly jig 6 to obtain the edge image of the assembled finished product (mainly the edge image of the sealing ring), and the second light source 32 is used to provide illumination for the second deep learning camera 31, so that the imaging of the second deep learning camera 31 is clearer.

[0039] In summary of the above technical description, the device of the present application integrates multiple functions for improving the quality of sealing ring assembly, such as deep image recognition, magnetic levitation non-damping conveying, vibration paving and picking trajectory correction, and adopts a mirror double station structure to improve the production efficiency. Compared with the prior art, the device of the present application can realize high-quality and high-efficiency electronic cigarette sealing ring assembly.

[0040] Embodiment two As shown in Figure 3 the embodiment discloses a sealing ring flexible assembly method based on mechanical vision technology, which is used for the sealing ring flexible assembly device based on mechanical vision technology described in embodiment one, The embodiment adopts a centralized controller to drive, control and transmit information of each electronic / power mechanism. Specifically, the method of the embodiment includes: S10: issuing an identification instruction to the first deep learning shooting module to obtain first coordinate data of the sealing ring.

[0041] Specifically, the first deep learning shooting module establishes an X-Y two-dimensional coordinate system on the vibration disc, and gives each sealing ring located inside the vibration disc a two-dimensional coordinate as the first coordinate data. Generally, the sealing ring is O-shaped. After taking the edge (taking the circumference of the circle), the first deep learning shooting module can calculate the two-dimensional coordinate of the center position of the circle through the centroid algorithm.

[0042] S20: predicting a first pick-and-place trajectory of the four-axis robot according to the first coordinate data and the current position of the four-axis robot, and sending an execution instruction of the first pick-and-place trajectory to the four-axis robot.

[0043] It should be explained that in the traditional pick-up process of the sealing ring, the initial position of the sealing ring needs to be specifically positioned. The step S20 mainly corrects the first pick-and-place trajectory of the four-axis robot according to the first coordinate data and the current position of the four-axis robot, so that the four-axis robot can adapt to the initial position of the sealing ring, pick up the sealing ring first, and then place it on the secondary positioning platform to realize automatic positioning of the sealing ring (without manual positioning operation).

[0044] Preferably, the prediction of the first pick-and-place trajectory can be based on a PID algorithm and a preset interference constraint (maximum allowed movement range of the shaft).

[0045] S30: sending a jig assembly instruction to the four-axis robot in response to completion of the execution instruction.

[0046] Specifically, after the execution instruction is completed, the sealing ring is automatically positioned again, and then the jig assembly instruction is sent to the four-axis robot, so that the four-axis robot performs jig assembly. The jig assembly instruction includes a cartridge-tobacco rod-sealing assembly instruction, a tank sealing assembly instruction, and / or an atomizer sealing assembly instruction. It should be explained that different assembly instructions have different opening degrees of the flexible clamp for supporting the sealing ring and different depths of the flexible clamp extending to the corresponding parts. Through the step S30, the method of the embodiment supports sealing ring assembly of various electronic cigarette parts.

[0047] S40: sending a snapshot instruction to the second deep learning shooting module in response to completion of the jig assembly instruction, and obtaining a snapshot image.

[0048] It should be explained that the snapshot technology is a technology for realizing efficient and accurate image acquisition and processing in a high-speed motion scene in the industrial vision field. The snapshot image obtained by the technology is essentially a shooting image of an object in motion. Through the step S40, the method of the application supports photographing of the assembled jig while conveying, thereby improving the production efficiency.

[0049] S50: comparing the snapshot image with a standard template image according to a preset deep learning algorithm, and calculating a matching degree of the snapshot image and the standard template image.

[0050] In the embodiment, the deep learning algorithm adopts a convolutional neural network model (CNN) or an adversarial network model (GAN). The step S50 includes: S51: Collect the edge features of the snapshot image to obtain a first edge feature matrix.

[0051] S52: Collect the edge features of the standard template image to obtain a second edge feature matrix.

[0052] S53: Calculate the cosine similarity of the first edge feature matrix and the second edge feature matrix, and take the cosine similarity as the matching degree.

[0053] More specifically, the plurality of standard template images are fed into the convolutional neural network model for deep learning and training, so that the convolutional neural network model has the basic judgment ability of electronic cigarette sealing assembly defects and obtains the second edge feature matrix. When the snapshot image is collected, the edge features of the snapshot image are extracted by using the convolutional neural network model to obtain the first edge feature matrix. Finally, the cosine similarity algorithm is used to calculate the similarity of the first edge feature matrix and the second edge feature matrix as the matching degree.

[0054] S60: If the matching degree is lower than the threshold, the assembled finished product is determined as a defective product.

[0055] In the embodiment, the threshold can be set to 98%.

[0056] Through the above steps S10-S60, the method of the embodiment realizes the function of artificial intelligence recognition assisting electronic cigarette sealing ring assembly, and further improves the output quality of electronic cigarette sealing ring assembly.

[0057] Preferably, before the above step S10, the method of the embodiment further comprises: sending a vibration instruction to the vibration disc for accommodating the sealing ring.

[0058] Through the above technical solution, the method of the embodiment can make the plurality of sealing rings horizontally laid on the vibration disc, thereby avoiding the four-axis robot taking multiple upper and lower overlapping sealing rings at a time. This way is conducive to further improving the output quality of electronic cigarette sealing ring assembly.

[0059] Although the present specification has shown and described several embodiments of the present application, it will be apparent to those skilled in the art that such embodiments are provided only by way of example. Those skilled in the art will think of many changes, changes and alternatives without departing from the idea and spirit of the present application. It should be understood that various alternatives to the embodiments of the application described herein can be employed in practicing the present application.

Claims

1. A mechanical vision technology-based flexible assembly equipment for sealing rings, characterized in that, The device comprises a vibrating disc, a first deep learning shooting module, a second deep learning shooting module, a four-axis robot, a transfer platform, and an assembly jig. The vibrating disc is used to accommodate the sealing ring to be assembled. The first deep learning shooting module is arranged on the top side of the vibrating disc and is used to identify the first coordinate data of the sealing ring and send the first coordinate data to the four-axis robot. The four-axis robot is used to pick up the sealing ring from the vibrating disc according to the first coordinate data and transport the sealing ring to the transfer platform for secondary positioning, and then transport the sealing ring to the base of the assembly jig for assembly after the secondary positioning is completed. The second deep learning shooting module is arranged on the top side of the assembly jig and is used to perform a snapshot detection on the assembled product in the assembly jig according to a preset detection algorithm, and send a rejection instruction to the master control device of the next workstation if the assembled product is detected as a defective product.

2. The mechanical vision technology based flexible assembly of seal ring apparatus as claimed in claim 1 wherein, The first deep learning shooting module comprises a first deep learning camera and a first light source, the first light source is arranged on the side of the first deep learning camera, and the lens of the first deep learning camera and the first light source are both vertically directed towards the vibrating disc.

3. The mechanical vision technology based flexible assembly of seal ring apparatus as claimed in claim 1 wherein, The second deep learning shooting module comprises a second deep learning camera and a second light source, the lens of the second deep learning camera and the second light source are both obliquely directed towards the assembly jig.

4. The mechanical vision technology based flexible assembly of seal ring apparatus as claimed in claim 1 wherein, The device further comprises a magnetic levitation guide rail, and the assembly jig is arranged on the top of the magnetic levitation guide rail.

5. The mechanical vision technology based flexible assembly of seal ring apparatus as claimed in claim 1 wherein, The vibrating disc is a flexible vibrating disc.

6. A method for flexible assembly of a sealing ring based on mechanical vision technology, characterized in that, The method for the flexible sealing ring assembly device based on mechanical vision technology according to any one of claims 1-5 comprises: sending an identification instruction to the first deep learning shooting module to obtain the first coordinate data of the sealing ring; predicting a first pick-and-place trajectory of the four-axis robot according to the first coordinate data and the current position of the four-axis robot, and sending an execution instruction of the first pick-and-place trajectory to the four-axis robot; sending a jig assembly instruction to the four-axis robot in response to the completion of the execution instruction; sending a snapshot instruction to the second deep learning shooting module in response to the completion of the jig assembly instruction to obtain a snapshot image; comparing the snapshot image with a standard template image according to a preset deep learning algorithm to calculate a matching degree of the snapshot image and the standard template image; determining that the assembled product is a defective product if the matching degree is lower than a threshold.

7. The method of claim 6, wherein the method further comprises: The deep learning algorithm adopts a convolutional neural network model or an adversarial network model.

8. The method of claim 6, wherein the method further comprises: The comparison of the snapshot image with the standard template image comprises: collecting edge features of the snapshot image to obtain a first edge feature matrix; collecting edge features of the standard template image to obtain a second edge feature matrix; calculating a cosine similarity of the first edge feature matrix and the second edge feature matrix, and taking the cosine similarity as the matching degree.

9. The method of claim 6, wherein the method further comprises: The jig assembly instruction comprises a cartridge-tobacco rod-sealing assembly instruction, an oil reservoir sealing assembly instruction, and / or an atomizer sealing assembly instruction.

10. The method of claim 6, wherein the method further comprises: Before sending the identification instruction to the first deep learning shooting module, the method further comprises: A vibration command is sent to a vibrating tray for receiving the sealing rings, so as to make a plurality of sealing rings horizontally spread on the vibrating tray.

Images