A fastening system and method based on machine vision and adaptive alignment

Abstract

The application relates to a fastening connection system and method based on machine vision and adaptive centering, and belongs to the technical field of intelligent manufacturing and assembly of large aerospace structures. Four-degree-of-freedom attitude adjustment is realized through an XY-axis displacement table, a lifting table and a rotating table. In view of the fact that the manual or mechanical centering mode is difficult to realize precise axis centering and multi-degree-of-freedom pose regulation and control due to the large size and heavy weight of large-size cylinder segments, the application realizes high-precision automatic centering and multi-degree-of-freedom pose dynamic regulation and control of the cylinder segments through multi-source vision real-time sensing and closed-loop feedback control. In view of the process defects such as local stress concentration, structural connection failure and sealing performance decline caused by sequential tightening of each bolt during multi-bolt fastening connection, the application realizes hierarchical progressive control through multi-bolt hierarchical collaborative loading and partition gradient strategy, and realizes real-time measurement of the pretightening force, tightening torque and tightening coefficient of each bolt, so that stress uniform distribution and full data traceability in the installation process are ensured.

Description

Technical Field

[0001] This invention belongs to the field of intelligent manufacturing and assembly technology for large aerospace structural components, specifically relating to a fastening connection system and method based on machine vision and adaptive alignment. Background Technology

[0002] In the aerospace field, component assembly often involves large cylindrical sections with numerous bolt holes and hundreds of bolts. Installation and alignment require significant labor costs, and the coordinated tightening of multiple bolts is also a major challenge. Furthermore, the lack of real-time position and orientation feedback during the tightening process and poor repeatability lead to structural deformation and uneven preload caused by improper assembly of large cylindrical sections and multiple bolts, posing potential risks to the reliability of spacecraft in service.

[0003] For the visual positioning of screw holes, the current main method is to use laser tracking and CCD visual recognition in the large component docking system of Star Company. The positioning accuracy is high, but the cost is high and it depends on a stable light source. The false detection rate increases in the case of curved surface reflection of the cylinder section (Zhu Yongguo et al., Dynamic comprehensive correction of aircraft wing-body docking assembly deviation based on laser tracker and machine vision, Journal of Mechanical Engineering, 2019, 55(24):10). In addition, ring cameras are used in the assembly of commercial aero-engines, but there is a lack of real-time pose feedback and cylinder section pose adjustment.

[0004] For multi-bolt tightening technology, Shenyang Aircraft Corporation of China (SAC) adopted multi-axis synchronous tightening in the assembly of the AG600 fuselage, but it can only achieve single torque value loading and lacks stress balance control, which leads to seal failure. In addition, there is an air-floating docking platform that solves the assembly friction problem, but it does not integrate bolt loading function and relies on manual intervention (a method for determining assembly stress of bolt fastener connection structure, CN115031873B; a fully automatic assembly docking method based on air-floating technology, CN106862908B).

[0005] For digital assembly platform technology, existing process simulation technology is disconnected from on-site execution, making it difficult to assemble large-sized cylindrical sections with multiple bolts in practice. Modular quick-release structures can achieve quick disassembly, but this is limited to small structures and does not solve the problem of deformation during assembly of large-sized cylindrical sections. Existing technologies mainly have the following problems: bolt hole visual positioning technology: poor environmental adaptability, insufficient real-time performance, and inability to achieve real-time feedback and adjustment of the position of large-sized cylindrical sections; multi-bolt tightening technology: coarse torque control, lack of zoned gradient strategy, and difficulty in achieving graded loading of "pre-tightening-equalization-final tightening"; single assembly platform technology: lack of reconfigurable modular design to solve the assembly scalability problem of large-sized cylindrical sections. Summary of the Invention

[0006] The purpose of this invention is to provide a fastening connection system and method based on machine vision and adaptive alignment. Through multi-dimensional pose control and multi-bolt hierarchical collaborative loading, it solves problems such as low alignment accuracy, uneven stress distribution, lack of pose control and poor platform scalability in the assembly of large-size compartments with multiple bolts.

[0007] The above-mentioned objectives of the present invention are mainly achieved through the following technical solutions: A fastening connection system based on machine vision and adaptive alignment is used to assemble bolts connecting a lower cylinder section and an upper cylinder section. The system includes an optical platform, a rotary platform, a first bearing platform, an XY-axis displacement platform, a second bearing platform, a lifting platform, a rotary table, a lifting device, and a robotic arm tightening unit. The rotary platform is mounted on the optical platform, and the first bearing platform is mounted on the rotary platform, rotating the first bearing platform. The XY-axis displacement platform is mounted on the first bearing platform, and the second bearing platform is mounted on the XY-axis displacement platform. The lower cylinder section is mounted on the second bearing platform, translating via the XY-axis displacement platform. The lifting platform is fixedly connected to the first bearing platform, and the rotary table is fixedly connected to the lifting platform. The upper cylinder section is mounted on the rotary table via the lifting device, and the upper cylinder section is raised and lowered via the lifting platform and rotated via the rotary table. The bolts are tightened by the robotic arm tightening unit.

[0008] The robotic arm tightening unit includes a servo motor, a torque sensor, and a bolt sleeve. The servo motor provides tightening power, the torque sensor records the tightening torque change curve in real time, and the bolt sleeve completes the tightening.

[0009] It also includes a ring array vision camera and a beam splitter. The ring array vision camera is mounted on the lifting platform and is used to acquire images of the bolt holes in the lower and upper cylinder sections. The beam splitter is used to assist the ring array vision camera in imaging. The focal point of the ring array vision camera and the center point of the beam splitter are on the same horizontal line.

[0010] After the bolt holes of the lower and upper cylinder sections are located by using a ring array vision camera and a beam splitter, the beam splitter retracts to prevent interference with the assembly of the lower and upper cylinder sections. The ring array vision camera monitors the assembly process in real time.

[0011] The optical platform, rotary platform, first bearing stage, XY axis displacement stage, and second bearing stage are detachable modular units, connected by standardized mechanical interfaces and quick-change connection structures.

[0012] It also includes an attitude adjustment control system, which drives the XY-axis displacement stage, lifting stage and rotary stage to achieve four-degree-of-freedom attitude adjustment.

[0013] A fastening connection method based on machine vision and adaptive alignment, using the above-mentioned assembly system, specifically includes the following steps: (1) Control the lifting platform to position the upper cylinder section at a safe height from the lower cylinder section; (2) Adjust the XY axis displacement table, lifting table and rotary table according to the bolt hole positions of the lower and upper cylinder sections. After alignment, control the lifting table to descend, place the upper cylinder section on the lower cylinder section, and disassemble the lifting device. (3) Control the tightening unit of the robotic arm to tighten.

[0014] In step (2), a visual processing algorithm is used to calculate the coordinates of the hole center and the cylinder center, and compare them with the target pose array to make a deviation judgment. After the pose adjustment is made according to the deviation judgment, a review and evaluation is performed. If the review and evaluation fails, the deviation judgment and pose adjustment are repeated until the review and evaluation passes.

[0015] In step (3), a multi-stage progressive method is used for tightening, as follows: (31) The robotic arm tightening unit applies initial torque to all bolts in sequence according to the set order, so that the bolts are initially attached and the initial preload is established; (32) Divide all bolts into bolt groups according to the area, apply intermediate gradient torque in the zone according to the set order, gradually equalize the distribution of bolt preload, and reduce the dispersion of preload; (33) Tighten the bolts to the target torque value to ensure that all bolts meet the design preload requirements.

[0016] An ultrasonic strain gauge is integrated at the end of the bolt to monitor the change in preload during tightening in real time. By synchronously acquiring the tightening torque T and preload F, and combining them with the nominal diameter d of the bolt, the tightening coefficient K of each bolt is calculated and recorded in real time. Based on the tightening coefficient K, the preload F is adjusted in real time.

[0017] Compared with the prior art, the present invention has at least the following beneficial effects: (1) This invention proposes a reconfigurable modular platform architecture for large-size sections. Unlike quick-disassembly solutions that are only applicable to small structural components, this invention addresses the special assembly requirements of large-size cylindrical sections by standardizing and modularizing core functional units such as optical platforms, rotary platforms, bearing platforms, and displacement platforms. The platform layout can be quickly reconfigured according to changes in the diameter and length of the cylindrical section, enabling one platform to cover the assembly tasks of multiple product models. This fundamentally solves the industry problems of high repetitive development costs and long changeover cycles of traditional dedicated assembly equipment.

[0018] (2) In the preferred embodiment of the present invention, a closed-loop adaptive centering system of "visual perception → pose calculation → multi-axis drive" is constructed. Unlike traditional open-loop positioning or single visual recognition schemes, the present invention deeply couples the ring array visual camera, the visual feedback system and the multi-degree-of-freedom posture adjustment mechanism to form a complete closed-loop control loop, realizing high-precision automatic centering and dynamic pose control of large-size cylindrical sections in multiple degrees of freedom such as X, Y, Z translation and rotation around the Z axis. The recognition and positioning accuracy reaches the level of 0.01mm, fundamentally solving the technical pain points of low accuracy and poor consistency of manual centering of large-size cylindrical sections.

[0019] (3) The preferred embodiment of the present invention establishes a multi-bolt "pre-tightening → equalization → final tightening" graded collaborative loading strategy and real-time mechanical sensing capability. The present invention breaks through the traditional extensive mode of single torque value loading. Through multiple sets of independently servo-driven robotic arm tightening units, combined with a partitioned gradient strategy and multi-level progressive loading process, the stress uniform distribution of hundreds of bolts in the entire assembly process is achieved. More importantly, through the collaborative work of torque sensors and ultrasonic strain gauges, the system can synchronously and in real time measure the pre-tightening force, tightening torque dynamic change curve and tightening coefficient value of each bolt, and construct a complete "torque-pre-tightening force-tightening coefficient" three-parameter assembly process data traceability chain, filling the technical gap of "unmeasurable, uncontrollable and untraceable" key mechanical parameters in the current assembly process of large-size cylindrical sections with multiple bolts.

[0020] (4) The preferred embodiment of the present invention realizes the coordinated perception and dynamic control of multiple physical quantities of "force-position-shape" in the assembly process. The present invention incorporates the mechanical parameters (preload, torque, tightening coefficient) and structural deformation parameters (radial deformation of the cylinder segment) in the tightening process into a unified real-time monitoring and feedback control framework. The tightening strategy is dynamically corrected with deformation perception data, and the torque output is controlled in a closed loop with mechanical feedback data, forming an intelligent decision-making closed loop of "perception → judgment → adjustment", which suppresses the risk of structural deformation and connection failure caused by uneven assembly stress in large-size thin-walled cylinder segments from the source. Attached Figure Description

[0021] Figure 1 This is an isometric view of the fastening connection system based on machine vision and adaptive alignment according to the present invention. Figure 2 This is a front view of the fastening connection system based on machine vision and adaptive alignment according to the present invention. Detailed Implementation

[0022] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments: This invention addresses the bottleneck of achieving precise axis alignment and multi-degree-of-freedom pose control through manual or mechanical alignment methods for large-sized cylindrical sections due to their large size and heavy weight (up to several tons). This invention achieves high-precision automatic alignment and dynamic multi-degree-of-freedom pose control of the cylindrical section through multi-source vision real-time perception and closed-loop feedback control.

[0023] Secondly, in response to the process defects such as local stress concentration, structural connection failure and reduced sealing performance caused by sequential tightening of multiple bolts during fastening, this invention achieves graded progressive control of the entire process of "pre-tightening → equalization → final tightening" of the bolt group through a multi-bolt graded collaborative loading and zoned gradient strategy. It also measures the pre-tightening force, tightening torque and tightening coefficient of each bolt in real time to ensure uniform stress distribution and full data traceability during the installation process.

[0024] Third, addressing the industry pain point that existing assembly equipment models are fixed and difficult to adapt to the assembly needs of compartments with different diameters and lengths, this invention adopts a reconfigurable modular architecture, which can quickly reconfigure the platform layout according to the external dimensions of the compartments, significantly improving the flexibility and versatility of the assembly system and reducing the cost of repeated development of multiple equipment models.

[0025] Fourth, in response to the safety hazard of structural deformation of large-sized thin-walled cylinder sections under assembly force, this invention integrates a real-time cylinder section deformation monitoring function during the tightening process. The tightening strategy is dynamically corrected based on deformation sensing data, thereby reducing the risk of structural failure induced by uneven assembly stress from the source.

[0026] like Figure 1 and Figure 2 As shown, the assembly platform of the present invention mainly includes: an optical platform 1, a rotary platform 2, a first bearing platform 3, an XY axis displacement platform 4, a second bearing platform 5, a lower cylinder section 6, an upper cylinder section 7, a lifting platform 8, a rotary table 9, a lifting device 10, a first robotic arm 11, a second robotic arm 12, a first servo motor 13, a second servo motor 14, a first torque sensor 15, a second torque sensor 16, a first bolt sleeve 17, a second bolt sleeve 18, a bolt 19, a ring array vision camera 20, and a beam splitter prism 21.

[0027] The connection relationships and collaborative working mechanisms between the components are as follows: The rotary platform 2 is mounted on the optical platform 1, and the first support platform 3 is mounted on the rotary platform 2. The rotary platform 2 drives the first support platform 3 to achieve 360° continuous rotation. The XY-axis displacement platform 4 is mounted on the support platform 3, and the second support platform 5 is mounted on the XY-axis displacement platform 4. The lower cylinder section 6 is fixed on the second support platform 5 and can move precisely in the XY direction with the XY-axis displacement platform 4. The lifting platform 8 passes through the through-hole in the middle of the XY-axis displacement platform 4 and the second support platform 5 and is directly fixed to the first support platform 3. The rotating platform 9 is directly connected to the lifting platform 8 and can move up and down in the Z direction with the lifting platform 8. The lifting device 10 lifts the upper cylinder section 7 and installs it on the rotating platform 9. The lifting platform 8 and the rotating platform 9 achieve Z-axis lifting and rotation around the Z-axis; the first robotic arm 11, the second robotic arm 12, the first servo motor 13, the second servo motor 14, the first torque sensor 15, the second torque sensor 16, the first bolt sleeve 17 and the second bolt sleeve 18 are assembled to form two independent robotic arm tightening units, which are distributed on both sides of the cylinder section; the ring array vision camera 20 is installed on the lifting platform 8, and the beam splitter prism 21 is directly connected to the vision camera system to ensure that the camera focus and the center point of the beam splitter prism are on the same horizontal line, which can not only photograph the screw hole positions of the upper cylinder section 6 and the lower cylinder section 7, but also monitor the external structure of the cylinder section by adjusting the focus.

[0028] (ii) Reconfigurable modular architecture The optical platform 1, rotary platform 2, first bearing platform 3, XY axis displacement platform 4, and second bearing platform 5 are all independent, detachable, modular units. The modules adopt standardized mechanical interfaces and quick-change connection structures, which can quickly replace and reassemble the modules according to the actual diameter and length of the cylindrical section to be assembled. This enables rapid switching from one size specification to another, fundamentally solving the scalability bottleneck of traditional fixed assembly platforms when facing multiple models of tasks. This allows a single platform to cover the assembly needs of compartments with multiple diameters and lengths.

[0029] (III) Multi-degree-of-freedom precision pose control mechanism The rotary platform 2 has a rotational stroke of 360° and is equipped with a rotary brake and locking device. Its repeatability is better than 0.001°, which can ensure the precise positioning and locking of the assembly platform at any angle.

[0030] The XY-axis displacement stage 4, lifting stage 8, and rotary stage 9 together form a multi-degree-of-freedom attitude adjustment mechanism, which is uniformly driven by the attitude adjustment control system. The XY-axis displacement stage 4 has a travel of 0-100mm and a positioning accuracy better than 0.01mm; the lifting stage 8 has a travel of 0-50mm and a positioning accuracy better than 0.01mm; and the rotary stage 9 has a travel of 0-360° and a positioning accuracy better than 0.001°. Through the coordinated movement of the three, precise attitude adjustment between the upper and lower cylinder sections in four degrees of freedom—X, Y, Z translation and rotation around the Z-axis—can be achieved.

[0031] The XY-axis displacement platform 4 and the second support platform 5 have a central opening structure and are fixedly connected to the first support platform 3. The lifting platform 8 can also be directly fixedly connected to the support platform 3, achieving decoupled and independent control of the XY translational motion of the lower cylinder section 6 and the Z-axis lifting and rotational motion of the upper cylinder section 7, ensuring that the attitude adjustments of the upper and lower cylinder sections do not interfere with each other. The XY-axis displacement platform can be replaced according to the external dimensions of the lower cylinder section 6 and the upper cylinder section 7.

[0032] (iv) Machine vision-guided adaptive centering closed-loop system The circular array vision camera 20, the vision camera feedback system, and the attitude adjustment control system together constitute a machine vision-guided cylindrical section adaptive centering closed-loop system. Its automatic centering process is as follows: (1) The circular array vision camera acquires images of bolt holes in the upper and lower cylinder sections. The vision processing unit performs image enhancement, edge detection and feature extraction, and calculates the center coordinates of the bolt holes and the center coordinates of the cylinder sections. (2) Compare the measured center coordinates of the hole with the preset target pose matrix and calculate the coordinate deviation; (3) The visual camera feedback system transmits the deviation judgment results to the attitude adjustment control system in real time; (4) The attitude adjustment control system calculates the compensation amount of each drive shaft according to the deviation value, and issues instructions to drive the XY displacement table and rotary table to perform platform compensation movement until the cylinder section and screw hole reach the target centering accuracy or preset assembly deviation tolerance. (5) The above identification → calculation → driving → verification process constitutes a closed-loop iterative control. The ring array vision camera contains multiple cameras, which can simultaneously locate multiple screw holes in the upper and lower cylinder sections. The minimum assembly deviation that can be identified is 0.01mm.

[0033] The focal point of the ring array vision camera and the center point of the beam splitter are on the same horizontal line, and the two beams on the beam splitter are vertical. After the ring array vision camera is fixed to the rotating stage, its installation position is flush with the top surface of the lower cylinder section, ensuring that the vision camera can clearly observe the position of the screw hole on the mating surface of the upper and lower cylinder sections.

[0034] After precise positioning is completed, the extended beam splitter prism can automatically retract to avoid interfering with subsequent assembly operations of the upper and lower cylinder sections. After the beam splitter prism retracts, the ring array vision camera can still continuously monitor the deformation of the inner diameter of the cylinder section in real time, with a minimum identifiable deformation of 0.1 mm, providing visual data support for monitoring structural deformation during the tightening process.

[0035] (V) Multi-bolt graded collaborative loading and real-time mechanical sensing The system includes two sets of robotic arm tightening systems. The first robotic arm tightening system consists of a first servo motor 13, a first torque sensor 15, and a first bolt sleeve 17. The second robotic arm tightening system consists of a second servo motor 14, a second torque sensor 16, and a second bolt sleeve 18. The robotic arm tightening system is not limited to two sets of tightening components and can be expanded into multiple tightening systems distributed around the circumference of the cylinder section according to actual needs, to achieve synchronous and coordinated loading of multiple bolts.

[0036] The servo motor can output a rated torque of not less than 200 Nm and an adjustable speed of 0-100 rpm. The servo motor can be quickly replaced according to the bolt specifications and tightening torque requirements.

[0037] The torque sensor is used to measure the output torque in real time, with a range of 250 Nm and a measurement accuracy better than 0.1%. The torque sensor can also feed the measurement results back to the servo motor controller in real time, forming a closed-loop precise control of the torque output. Torque sensors with different ranges and accuracies can be replaced according to actual measurement needs. The bolt sleeve transmits torque to the bolt and can be quickly replaced according to the bolt specifications and structure.

[0038] Based on the aforementioned hardware, the servo motor can output different tightening torques, tightening speeds, and tightening angles in segments, thereby realizing a multi-stage progressive tightening strategy of "pre-tightening → equalization → final tightening". Specifically: Pre-tightening stage: Each robotic arm applies an initial small torque to the bolts in a predetermined sequence to make the bolts initially fit together and establish the initial pre-tightening force; Equilibrium Phase: Apply intermediate gradient torque to each bolt group according to the zoning strategy to gradually equalize the preload distribution of each bolt and reduce the preload dispersion between bolts; Final tightening stage: Complete the final tightening according to the target torque value to ensure that all bolts meet the design preload requirements.

[0039] This invention features real-time measurement functions for preload, tightening torque, and tightening coefficient: During the tightening process, as the tightening system applies the target torque to each fastener, the dynamic change curve of the tightening torque is recorded in real time by the first torque sensor 15 and the second torque sensor 16. Simultaneously, the ultrasonic strain gauge integrated at the end of the bolt 19 monitors the preload change of each fastener in real time through an ultrasonic stress measurement system. By synchronously acquiring the tightening torque T and preload F, combined with the nominal bolt diameter d, the tightening coefficient K of each bolt is calculated and recorded in real time (K = T / (F•d)), forming a complete three-parameter assembly process data chain of "torque-preload-tightening coefficient," providing comprehensive quantitative basis for assembly quality assessment and process optimization.

[0040] The rotary platform 2 is fixedly connected to the attitude adjustment mechanism. By rotating the rotary platform 2, the lower cylinder section 6 and the upper cylinder section 7 can be rotated as a whole, and the bolts 19 to be tightened and the screw holes can be positioned sequentially or in groups to the corresponding working positions of the robotic arm tightening system, thereby supporting the coordinated tightening of multiple groups of bolts at different points.

[0041] (vi) Complete assembly process The aerospace large-size cylindrical section multi-bolt assembly platform device of the present invention includes the following assembly method: Step 1: Sequentially assemble the rotary platform 2, the first bearing platform 3, the XY axis displacement platform 4, the second bearing platform 5, the lifting platform 8, the rotating platform 9, and the ring array vision camera 20 on the optical platform to form the main structure of the cylindrical segment multi-bolt assembly platform. Each module is selected and combined according to the size of the cylindrical segment to be assembled. Step 2: Fix the lower cylinder section 6 onto the second bearing platform 5, ensuring that the lower cylinder section 6 is aligned with and concentric with the center of the main structure of the multi-bolt assembly platform; Step 3: Connect the upper cylinder section 7 to the lifting device 10 with bolts, and install the upper cylinder section 7 on the rotary table 9 through the lifting device 10; Step 4: Control the lifting platform 8 to position the upper cylinder section 7 at a certain safe height from the lower cylinder section 6, start the ring array vision camera 20 to photograph the bolt holes and outer wall of the upper and lower cylinder sections, calculate the center coordinates of the holes and the center of the cylinder section through the vision processing algorithm, compare them with the target pose array, and make a deviation judgment. Step 5: Based on the coordinate deviation value, the vision camera feedback system transmits the deviation information to the pose adjustment control system. After the system calculates the deviation, it issues drive commands to the XY displacement stage and rotary stage, driving the platform to move to the target coordinates. The circular array vision camera is then used for verification and evaluation to confirm whether the predetermined assembly position has been reached. Steps 4-5 above constitute a complete closed-loop adaptive alignment process of "visual recognition → deviation calculation → pose adjustment → verification and confirmation". Step 6: The beam splitter prism 21 in the circular array vision camera system automatically retracts to ensure that the retracted beam splitter prism does not affect the subsequent docking and installation of the upper and lower cylinder sections. Step 7: Control the lifting platform to slowly descend, place the upper cylinder section smoothly on the lower cylinder section, and remove the lifting equipment; Step 8: Install multiple sets of bolts into the corresponding bolt holes. Using the robotic arm system, set the tightening parameters (torque, speed, angle) for each stage according to the preset tightening strategy, and start the graded loading program of "pre-tightening → equalization → final tightening" until all bolts reach the target tightening state. During the entire tightening process, the tightening torque change curve is recorded in real time by the torque sensor, and the pre-tightening force change of each bolt is measured in real time by the ultrasonic strain gauge and ultrasonic stress measurement system, and the tightening coefficient value is calculated accordingly. Step 9: During the tightening process, the outer structure of the cylinder section is monitored in real time by a ring array vision camera to identify the deformation of the cylinder section. The deformation data is fed back to the tightening control system in real time. The system dynamically adjusts the tightening parameters accordingly to reduce the risk of uneven stress caused by improper tightening strategy.

[0042] All components of the above assembly system adopt standardized interface design, which enables rapid expansion and reconfiguration for different cylinder section dimensions, significantly improving the system's assembly flexibility and adaptability to multiple models.

[0043] The above description is only the best specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the protection scope of the present invention.

[0044] The contents not described in detail in this specification are common knowledge to those skilled in the art.

Claims

1. A fastening connection system based on machine vision and adaptive alignment for assembling bolts (19) connecting a lower cylinder section (6) and an upper cylinder section (7), characterized in that: The system includes an optical platform (1), a rotary platform (2), a first support platform (3), an XY-axis displacement platform (4), a second support platform (5), a lifting platform (8), a rotary table (9), a lifting device (10), and a robotic arm tightening unit. The rotary platform (2) is mounted on the optical platform (1), the first support platform (3) is mounted on the rotary platform (2), and the rotary platform (2) drives the first support platform (3) to rotate. The XY-axis displacement platform (4) is mounted on the first support platform (3), and the second support platform (5) is mounted on the optical platform (1). On the XY axis displacement stage (4), the lower cylinder section (6) is installed on the second bearing stage (5). The lower cylinder section (6) is moved horizontally by the XY axis displacement stage (4). The lifting stage (8) is fixedly connected to the first bearing stage (3). The rotating stage (9) is fixedly connected to the lifting stage (8). The upper cylinder section (7) is installed on the rotating stage (9) by the hoist (10). The upper cylinder section (7) is lifted and lowered by the lifting stage (8). The upper cylinder section (7) is rotated by the rotating stage (9). The bolt (19) is tightened by the mechanical arm tightening unit.

2. The fastening connection system based on machine vision and adaptive alignment according to claim 1, characterized in that: The robotic arm tightening unit includes a servo motor, a torque sensor, and a bolt sleeve. The servo motor provides tightening power, the torque sensor records the tightening torque change curve in real time, and the bolt sleeve completes the tightening.

3. The fastening connection system based on machine vision and adaptive alignment according to claim 1, characterized in that: It also includes a ring array vision camera (20) and a beam splitter (21). The ring array vision camera (20) is mounted on the lifting platform (8) and is used to collect images of the bolt holes of the lower cylinder section (6) and the upper cylinder section (7). The beam splitter (21) is used to assist the ring array vision camera (20) in imaging. The focal point of the ring array vision camera (20) and the center point of the beam splitter (21) are on the same horizontal line.

4. The fastening connection system based on machine vision and adaptive alignment according to claim 3, characterized in that: After the bolt holes of the lower cylinder section (6) and the upper cylinder section (7) are positioned by the ring array vision camera (20) and the beam splitter (21), the beam splitter (21) retracts to prevent interference with the assembly of the lower cylinder section (6) and the upper cylinder section (7). The ring array vision camera (20) monitors the assembly process in real time.

5. A fastening connection system based on machine vision and adaptive alignment according to claim 1, characterized in that: The optical platform (1), rotary platform (2), first bearing platform (3), XY axis displacement platform (4), and second bearing platform (5) are detachable modular units, connected by standardized mechanical interfaces and quick-change connection structures.

6. The fastening connection system based on machine vision and adaptive alignment according to claim 1, characterized in that: It also includes an attitude adjustment control system, which drives the XY-axis displacement stage (4), the lifting stage (8) and the rotary stage (9) to achieve four-degree-of-freedom attitude adjustment.

7. A fastening connection method based on machine vision and adaptive alignment, characterized in that: The assembly system according to any one of claims 1 to 6 specifically includes the following steps: (1) Control the lifting platform (8) to position the upper cylinder section (7) at a safe height from the lower cylinder section (6); (2) Adjust the XY axis displacement table (4), lifting table (8) and rotating table (9) according to the bolt hole positions of the lower cylinder section (6) and the upper cylinder section (7). After alignment, control the lifting table (8) to descend, place the upper cylinder section (7) on the lower cylinder section (6), and disassemble the lifting device (10). (3) Control the tightening unit of the robotic arm to tighten.

8. The fastening connection method based on machine vision and adaptive alignment according to claim 7, characterized in that: In step (2), a visual processing algorithm is used to calculate the coordinates of the hole center and the cylinder center, and compare them with the target pose array to make a deviation judgment. After the pose adjustment is made according to the deviation judgment, a review and evaluation is performed. If the review and evaluation fails, the deviation judgment and pose adjustment are repeated until the review and evaluation passes.

9. A fastening connection method based on machine vision and adaptive alignment according to claim 7, characterized in that: In step (3), a multi-stage progressive method is used for tightening, as follows: (31) The mechanical arm tightening unit applies initial torque to all bolts (19) in sequence according to the set order, so that the bolts (19) are initially attached and the initial preload is established; (32) Divide all bolts (19) into bolt groups according to the area, apply intermediate gradient torque in the partition according to the set order, gradually equalize the distribution of bolt (19) preload, and reduce the preload dispersion; (33) Tighten the bolts to the target torque value to ensure that all bolts (19) meet the design preload requirements.

10. A fastening connection method based on machine vision and adaptive alignment according to claim 9, characterized in that: An ultrasonic strain gauge is integrated at the end of the bolt (19) to monitor the change of preload during the tightening process in real time. By synchronously acquiring the tightening torque T and preload F, combined with the nominal diameter d of the bolt (19), the tightening coefficient K of each bolt (19) is calculated and recorded in real time. Based on the tightening coefficient K, the preload F is adjusted in real time.

Images