Automatic assembling system and method for large nut of water turbine ball valve

Abstract

The application belongs to the technical field of automatic assembly, and discloses an automatic assembly system and method for large nuts of a water turbine ball valve, which comprises a mechanical arm, an end effector and an intelligent nut tightening control system. The end effector is integrated with an electromagnet adsorption mechanism, a servo-driven rotating mechanism, a floating compensation mechanism and a visual recognition mechanism. Different specifications of nuts are adaptively grabbed by the electromagnet, sub-millimeter-level accurate positioning is realized by using a laser galvanometer stereo camera, and the assembly posture is automatically adjusted in combination with the floating compensation mechanism. The intelligent control system adopts a phased tightening strategy, realizes real-time monitoring through a torque sensor, and has a double protection mechanism. The application realizes the full automation of the assembly of large nuts of a water turbine ball valve, has the advantages of high assembly efficiency, accurate positioning precision, strong adaptability, reliable quality and the like, solves the problems of high labor intensity, many safety hazards and poor quality consistency in manual assembly, and significantly improves the assembly quality and efficiency of key components of a hydropower station.

Description

Technical Field

[0001] This invention belongs to the field of automated assembly technology, and in particular relates to an automated assembly system and method for large nuts of water turbine ball valves. Background Technology

[0002] The assembly of turbine ball valves is a crucial step in the operation of hydropower stations, involving the installation and tightening of multiple high-strength nuts. These nuts are large in size and weight, and require precise pre-tightening during installation to ensure the sealing and safety of the ball valves. However, traditional manual assembly methods are not only labor-intensive but also pose safety hazards, easily leading to problems such as nut jamming and stripped threads, affecting assembly quality and efficiency. Especially for nuts weighing up to hundreds of kilograms, failure to accurately control the torque or tightening angle during manual operation can result in nut damage or assembly errors.

[0003] Furthermore, the complexity of ball valve design increases the difficulty of nut installation. Different sized nuts are installed on the four sides of the ball valve, and each nut has different preload and torque requirements, posing a significant challenge to manual assembly. With the increasing scale of hydropower station installations, traditional manual assembly methods can no longer meet the requirements of efficient, safe, and precise assembly. Therefore, there is an urgent need for an automated and intelligent assembly system to improve production efficiency, reduce human error, and ensure operational safety.

[0004] Existing automated assembly technologies are mostly applied to production lines for light workpieces. However, for the assembly of large and heavy equipment such as turbine ball valves, related automation technologies are still in the exploratory stage. Traditional robotic arms and vision systems face several challenges when handling precise nut installation tasks, such as accurate identification of nut positions and dynamic adjustments in complex environments. Furthermore, most existing assembly systems fail to achieve end-to-end intelligent control, lacking effective automated control programs and monitoring mechanisms, leading to errors and inefficiencies in practical applications.

[0005] Therefore, developing an efficient, safe, and intelligent turbine ball valve nut assembly system has become an urgent problem to be solved. Summary of the Invention

[0006] The purpose of this invention is to provide an automatic assembly system and method for large nuts of turbine ball valves to solve the above-mentioned technical problems.

[0007] To solve the above-mentioned technical problems, the specific technical solution of the automatic assembly system and method for large nuts of turbine ball valves of the present invention is as follows:

[0008] An automated assembly system for large nuts on a turbine ball valve includes a robotic arm, an end effector, and an intelligent nut tightening control system. The robotic arm is mounted at a designated working point on the turbine ball valve platform and is controlled by the intelligent nut tightening control system. The end effector is mounted at the end of the robotic arm and performs identification, gripping, and tightening operations. The intelligent nut tightening control system controls the operation of the robotic arm and the end effector. The end effector includes an electromagnet adsorption mechanism, a servo-driven rotation mechanism, a floating compensation mechanism, and a vision recognition mechanism. The servo-driven rotation mechanism and the vision recognition mechanism are mounted on the end effector. The electromagnet adsorption mechanism is mounted on the servo-driven rotation mechanism via the floating compensation mechanism. The electromagnet adsorption mechanism uses electromagnet adsorption to grip nuts of different specifications. The servo-driven rotation mechanism is used to tighten the nuts. The floating compensation mechanism automatically adjusts the electromagnet's posture by extending and retracting during the nut screwing process to compensate for assembly errors. The vision recognition mechanism is used to identify the position, geometric features, and three-dimensional coordinates of the assembly point of the turbine ball valve nut in real time.

[0009] Furthermore, the servo-driven rotation mechanism includes a mounting base, a servo motor, and a gripper. One end of the mounting base is fixedly connected to the end of the robotic arm, and the other end is equipped with the servo motor, gripper, and electromagnet adsorption mechanism. The servo motor is fixedly mounted on the upper end of the mounting base, and the gripper is mounted on the lower end of the mounting base. The gripper includes a base and a claw. The base is fixedly connected to the output end of the servo motor, and the claws are distributed around the base and fixedly connected to the base. A lever is installed on the side of the nut to be installed. The rotation of the servo motor drives the gripper to rotate, and the claws of the gripper move the lever to tighten the nut.

[0010] Furthermore, the electromagnet adsorption mechanism includes an electromagnet, which is installed below the base of the gripper via a floating compensation mechanism. The electromagnet dynamically adjusts the magnetic force according to the material and size of the nut.

[0011] Furthermore, the floating compensation mechanism includes multiple long screws, which are evenly distributed on the base of the gripper. The long screws are slidably connected to the base, and the upper end of each long screw has a nut to limit its extension length and ensure that it does not come off the base. The lower end of each long screw is fixedly connected to an electromagnet.

[0012] Furthermore, the visual recognition mechanism includes a laser galvanometer stereo camera, which is mounted in the center of the mounting base and used to identify the pose, geometric features, and three-dimensional coordinates of the assembly point of the turbine ball valve nut in real time. The laser galvanometer stereo camera includes a laser, a galvanometer system, and an image sensor. The laser emits a high-precision laser beam; the galvanometer system includes a high-speed scanning galvanometer or an optical microelectromechanical system (MEMS) galvanometer for adjusting the scanning direction of the laser beam; and the image sensor employs a high-sensitivity CMOS or CCD sensor to calculate the time of flight or angular deviation of the laser signal, thereby determining the depth information of the object's surface.

[0013] Furthermore, an electromagnetic torque sensor and a force sensor are installed on the end effector to monitor the torque and mechanical changes during the locking process in real time.

[0014] Furthermore, the intelligent nut tightening control system includes the following functional modules:

[0015] Image acquisition module: Acquires point cloud images of the working environment in real time through a laser galvanometer stereo camera;

[0016] Image processing module: Utilizing the Open3D point cloud processing library for data preprocessing, point cloud feature extraction, and target recognition, the image processing module includes a data preprocessing module, a feature extraction and matching module, a point cloud registration module, a result optimization and verification module, and a pose calculation module. The data preprocessing module filters, reduces noise, and removes outliers based on point cloud position and distance. The feature extraction and matching module performs initial estimation for point cloud registration. The point cloud registration module combines RANSAC and ICP algorithms to achieve accurate point cloud alignment. The result optimization and verification module performs error analysis on the registration results. The pose calculation module calculates the optimal grasping pose of the robotic arm using the SVD algorithm and transforms the visual coordinates to the robotic arm coordinate system.

[0017] Robotic arm control module: The robot arm acquires displacement and posture information in real time through sensors inside the robot arm and performs high-precision control, including path planning and real-time feedback. The path planning is based on the teaching trajectory learned by Dynamic MovementPrimitives and combined with obstacle avoidance algorithm to optimize the path. The joint angle is calculated by inverse kinematics to generate an efficient collision-free motion trajectory. The movement path of the robot arm (1) is arranged in a diagonal installation order so that the end effector (2) tightens each nut on the working surface one by one and repeats the operation twice. The real-time feedback dynamically adjusts the robot arm's movements through image data acquired by torque sensor and laser galvanometer stereo camera (27).

[0018] Torque control module: Prevents overtightening by monitoring torque changes in real time; sets different locking parameters according to different types of nuts and ball valves; adopts a staged tightening strategy: small torque for rapid tightening in the pre-tightening stage, and high-precision torque control in the final stage; integrates torque-angle method and yield point control method to optimize the torque curve; uses an electromagnetic torque sensor to monitor mechanical changes during the tightening process to ensure that each nut maintains stable pressure throughout the assembly process; optimizes fine-tuning during the assembly process through a force feedback mechanism.

[0019] Dual protection module: including force sensor monitoring module and vision monitoring module. The force sensor monitoring module includes a force sensor on the end effector (2) to detect the force on the nut in real time. When the friction or torque exceeds the set value, the system will automatically stop and issue an alarm to prompt the operator to check. The vision monitoring module includes the system's laser galvanometer stereo camera (27) to monitor the assembly process in real time and feed it back to the system program to ensure the correct position and orientation of the nut.

[0020] Data Management Module: Records the assembly data of each ball valve and maps this data to the unique identifier of each ball valve. The data management module supports cloud storage. The data management module includes a data recording and analysis module and a fault tolerance and recovery module. The data recording and analysis module records relevant data during the robot's operation. The fault tolerance and recovery module considers various possible error situations and has corresponding fault tolerance and recovery mechanisms.

[0021] This invention also discloses an automatic assembly method for an automatic assembly system of large nuts for turbine ball valves, comprising the following steps:

[0022] Step 1: Target detection and positioning: Based on the position and posture of the target output by the vision recognition mechanism, the system guides the robotic arm to perform precise assembly operations. At the same time, the working status and assembly progress of the robotic arm are displayed through a real-time monitoring interface. A laser beam is emitted by a laser galvanometer stereo camera, combined with a high-precision image sensor, to acquire the three-dimensional point cloud data of the workpiece in real time.

[0023] Step 2: Path Planning and Execution: Combining the data identified by the vision recognition mechanism, the system automatically generates the gripping and assembly paths for the robotic arm. The system instructs the robotic arm to move to the nut installation point along the optimal path through the control program. The movement path of the robotic arm follows a diagonal installation sequence, causing the end effector to tighten each nut on the working surface one by one, and repeating the operation twice. The DMP algorithm is used to optimize the path of the robotic arm to ensure that the path does not interfere with other equipment or obstacles.

[0024] Step 3: Assembly execution and control: The robotic arm grabs the nut and positions it to the nut through the end effector, and performs the tightening action, while the floating compensation mechanism compensates for the error;

[0025] Staged tightening strategy: small torque for rapid tightening in the pre-tightening stage, and high-precision torque control in the final stage to ensure proper tightening;

[0026] Step 4: Quality Monitoring and Feedback: Real-time Feedback and Adjustment: During the robotic arm's path planning process, the system receives sensor feedback data in real time and adjusts the posture to ensure precise alignment between the end effector 2 and the nut mounting point. During the nut tightening process, the force sensor monitors the applied force on the nut in real time. If the torque exceeds the set threshold, the system will immediately stop the operation and issue an alarm, prompting the operator to perform a manual inspection to prevent the nut from jamming or being damaged.

[0027] Furthermore, step 1 includes the following steps:

[0028] Step 1.1: Image acquisition and data preprocessing: Use a laser galvanometer stereo camera (27) to acquire three-dimensional point cloud data and image information in real time to accurately obtain the spatial coordinates and geometry of the nut installation area; Voxel filtering: downsample the original point cloud; Statistical filtering: remove outliers and optimize the point cloud data quality; Radius filtering: further remove isolated points;

[0029] Step 1.2: Image Processing and Feature Recognition: Point cloud feature extraction, coarse registration, and fine registration are used to identify the specific position and angle of the nut;

[0030] Step 1.3: Pose calculation: Based on the position and orientation of the nut identified by the vision system, calculate the spatial pose of the nut to guide the robotic arm (1) to accurately grasp and assemble it; through the transformation relationship between the vision coordinate system and the robotic arm coordinate system, transmit the nut installation point position information to the intelligent nut tightening control system;

[0031] 3D Reconstruction: Using the calibrated camera parameters, the 2D image coordinates of the nut are converted into 3D spatial coordinates, and point cloud data of the nut's mounting points are generated. Based on the camera's intrinsic and extrinsic parameters, the 2D image coordinates (u,v) are converted into 3D spatial coordinates (X,Y,Z) using the following formula:

[0032]

[0033] Where R is the rotation matrix and T is the translation vector. This transformation yields the precise position of the nut in three-dimensional space.

[0034] Pose determination: The optimal pose of the robotic arm when grasping the nut is calculated by using the singular value decomposition algorithm SVD. The best grasping posture of the nut is fitted using the feature points of the point cloud data.

[0035] Visual correction and position calibration: During the assembly process, the system continuously adjusts the posture of the robotic arm through visual feedback to correct deviations caused by nut position errors or angle deviations, ensuring that the nut is correctly assembled in place;

[0036] Foreign object detection in screw holes: Infrared or ultrasonic sensors are used to monitor whether there are foreign objects inside the assembly holes. When an abnormal distance or an obstacle is detected, the system immediately stops the assembly process and notifies the operator.

[0037] Furthermore, step 1.2 includes the following steps:

[0038] Step 1.2.1: Point cloud feature extraction: Calculate the normal vector information of the point cloud to provide a basis for subsequent feature matching;

[0039] Step 1.2.2: Coarse registration: Fast Point Feature Histogram (FPFH) is used to extract local geometric features of the point cloud to achieve preliminary point cloud matching; the initial transformation estimation is performed based on the Random Sample Consensus Algorithm (RANSAC) to obtain a relatively accurate rotation matrix and translation vector; the sampling consensus iterative registration algorithm (SAC-IA) is used to find the initial transformation matrix through feature matching.

[0040] Step 1.2.3: Fine registration: Iterative nearest point algorithm ICP is used: Based on the least squares optimization method, the point cloud is accurately registered; Weighted ICP: Based on the traditional ICP, the matching point pairs are assigned weights to improve the matching accuracy; Color ICP: RGB information is used to optimize the point cloud matching.

[0041] The automatic assembly system and method for a large nut on a turbine ball valve of the present invention has the following advantages:

[0042] 1. Compatible with various nut specifications, highly adaptable

[0043] Employing an electromagnet adsorption mechanism combined with a floating compensation design, the magnetic force can be dynamically adjusted to adapt to nuts of different sizes and materials, solving the problem of frequent changes required by traditional clamps and offering greater versatility. This significantly reduces manual operation and greatly improves assembly efficiency.

[0044] 2. High-precision assembly and error compensation

[0045] The visual recognition mechanism acquires 3D point cloud data in real time through a laser galvanometer stereo camera and achieves sub-millimeter positioning accuracy by combining algorithms such as RANSAC and ICP; the floating compensation mechanism automatically adjusts its posture during the tightening process to effectively compensate for assembly errors and ensure precise alignment of the nut and bolt.

[0046] 3. Intelligent tightening control and quality assurance

[0047] The phased tightening strategy (rapid tightening in the pre-tightening stage + high-precision torque control in the final tightening stage) combined with real-time monitoring by an electromagnetic torque sensor avoids over-tightening or loosening; the dual protection module (force sensor + visual monitoring) can immediately stop the machine and alarm in case of abnormality, eliminating the risk of thread damage or jamming.

[0048] 4. Data traceability and process optimization

[0049] The data management module records the assembly parameters (such as torque and sequence) of each nut and binds them to the unique identifier of the ball valve. It supports cloud storage and analysis, providing data support for quality traceability and process improvement.

[0050] 5. Adaptability to complex environments

[0051] The laser galvanometer stereo camera has strong resistance to ambient light interference and can work stably in complex scenarios such as the confined space of a water turbine ball valve and multiple reflective surfaces, ensuring reliable recognition.

[0052] In summary, this invention, through intelligent and modular design, achieves efficient, precise, safe, and traceable assembly of large nuts for turbine ball valves, providing an innovative solution for the automated assembly of hydropower equipment. Attached Figure Description

[0053] Figure 1 This is a schematic diagram showing the position of the automatic assembly system of the present invention installed on the ball valve of the water turbine;

[0054] Figure 2 This is a schematic diagram of the robotic arm structure of the present invention;

[0055] Figure 3 This is a schematic diagram of the end effector structure of the present invention;

[0056] Figure 4 This is a schematic diagram of the nut structure of the present invention;

[0057] Figure 5 This is a schematic diagram of the working process of the laser galvanometer stereo camera of the present invention;

[0058] Figure 6 This is a schematic diagram of the robotic arm's movement path according to the diagonal installation sequence of the present invention;

[0059] Figure 7 This is a torque control table for the phased tightening strategy of the present invention;

[0060] Figure 8 This is a flowchart of the nut position identification process of the present invention;

[0061] The markings in the diagram are as follows: 1. Robotic arm; 2. End effector; 21. Mounting base; 22. Servo motor; 23. Gripper; 231. Base; 232. Claw; 24. Lever; 25. Electromagnet; 26. Long screw; 261. Nut; 27. Laser galvanometer stereo camera. Detailed Implementation

[0062] To better understand the purpose, structure, and function of this invention, the following detailed description, in conjunction with the accompanying drawings, provides an automatic assembly system and method for a large nut on a water turbine ball valve.

[0063] like Figure 1 Figure 2 As shown, the present invention provides an automatic assembly system for a large nut of a turbine ball valve, comprising a robotic arm 1, an end effector 2, and an intelligent nut tightening control system. The robotic arm 2 is installed at a designated working point on the platform of the turbine ball valve and is controlled by the intelligent nut tightening control system. The end effector 2 is installed at the end of the robotic arm 1 and is used to perform identification, gripping, and tightening operations. The intelligent nut tightening control system is used to control the operation of the robotic arm 1 and the end effector 2.

[0064] like Figure 3 As shown, the present invention designs an end effector 2 specifically for assembling ball valve nuts in water turbines. The end effector 2 includes an electromagnet adsorption mechanism, a servo drive rotation mechanism, a floating compensation mechanism, and a vision recognition mechanism. The servo drive rotation mechanism and the vision recognition mechanism are mounted on the end effector 2, and the electromagnet adsorption mechanism is mounted on the servo drive rotation mechanism through the floating compensation mechanism.

[0065] The servo-driven rotary mechanism includes a mounting base 21, a servo motor 22, and a gripper 23. One end of the mounting base 21 is fixedly connected to the end of the robotic arm 1, and the other end is equipped with the servo motor 22, the gripper 23, and an electromagnet adsorption mechanism. The servo motor 22 is fixedly mounted on the upper end of the mounting base 21, and the gripper 23 is mounted on the lower end of the mounting base 22. The gripper 23 includes a base 231 and grippers 232. The base 231 is fixedly connected to the output end of the servo motor 22, and the grippers 232 are distributed around the base 231 and fixedly connected to the base 231. Figure 4 As shown, a lever 24 is installed on the side of the nut to be installed. The servo motor 22 rotates to drive the gripper 23 to rotate. The claw 232 of the gripper 23 moves the lever 24 to tighten the nut.

[0066] The electromagnet adsorption mechanism includes an electromagnet 25, which is mounted below the base 231 of the gripper 23 via a floating compensation mechanism. The electromagnet 25 is used to adsorb and grip different types of nuts. The electromagnet 25 can dynamically adjust its magnetic force according to the material and size of the nut, ensuring stable gripping and release of different types of nuts. Using electromagnet adsorption and gripping allows for compatibility with nuts of different specifications, enabling a single structure to grip nuts of different sizes, greatly improving production efficiency.

[0067] The floating compensation mechanism includes multiple long screws 26, which are evenly spaced on the base 231 of the gripper 23. The long screws 26 are slidably connected to the base 231, and each long screw 26 has a nut 261 at its upper end to limit its extension length and prevent it from detaching from the base 231. The lower end of each long screw 26 is fixedly connected to an electromagnet 25. This floating compensation mechanism allows the end effector 2 to automatically adjust the attitude of the electromagnet 25 during the screwing-in of the nut, compensating for assembly errors.

[0068] The visual recognition mechanism includes a laser galvanometer stereo camera 27, which is mounted in the middle of the mounting base 21. It is used to identify the pose, geometric features, and three-dimensional coordinates of the assembly point of the turbine ball valve nut in real time. The laser galvanometer stereo camera 27 includes a laser, a galvanometer system, and an image sensor. The laser emits a high-precision laser beam, typically a pulsed or continuous-wave laser, to adapt to different measurement scenarios. The laser wavelength is selected from the infrared or visible light range to ensure good reflectivity to different material surfaces. Preferably, the laser is a multi-wavelength laser, which improves measurement accuracy and anti-interference capability. The galvanometer system includes a high-speed scanning galvanometer or an optical microelectromechanical system (MEMS) galvanometer, which can quickly adjust the scanning direction of the laser beam. This invention uses dual-axis galvanometer control, enabling the laser to cover the entire surface of the object being measured according to a set scanning path. The image sensor uses a high-sensitivity CMOS or CCD sensor, combined with time-of-flight (ToF) measurement or triangulation, to calculate the flight time or angular deviation of the laser signal, thereby determining the depth information of the object's surface. Image sensors are equipped with filters to reduce ambient light interference and improve data quality.

[0069] The working principle of the laser galvanometer stereo camera 27 is as follows: Figure 5As shown, during the measurement process, a laser beam is emitted by a laser, and the direction of the laser beam is rapidly adjusted by a galvanometer system to scan the entire object under test. Simultaneously, an image sensor captures images of the object at a specified frame rate. Based on the captured image data, the galvanometer system calculates the three-dimensional coordinate information of the object's surface by analyzing the position and characteristics of the laser reflection points. Finally, combined with the calibration parameters of the galvanometer system, high-precision three-dimensional point cloud data is generated. This method can effectively acquire detailed depth information of complex scenes, effectively solving the problems caused by various adverse factors such as light absorption, high reflectivity, and ambient light interference, ensuring accurate, stable, and reliable 3D imaging of the workpiece.

[0070] The end effector 3 is controlled by a microcontroller, which interacts with the system controller. The controller sends commands to the microcontroller, enabling the microcontroller to control the actuator's response, including the on / off state of the electromagnet 25 and the rotation of the servo motor 22. The entire process involves the electromagnet 25 being energized to pick up the nut, and the robotic arm 2 moving the end effector 3 to the coordinates and orientation of the nut mounting point as determined by a vision recognition mechanism. The servo motor 22 then starts working, causing the electromagnet 25 and the nut to rotate counter-clockwise. Under the action of the floating compensation mechanism, the nut is rotated onto the bolt of the turbine ball valve. An electromagnetic torque sensor and a force sensor are installed on the end effector 2 to monitor the torque and mechanical changes during the tightening process in real time.

[0071] The intelligent nut tightening control system is based on a multi-sensor fusion control framework built on the TCP / IP protocol. It achieves closed-loop coordination of visual recognition, robotic arm control, and torque feedback, ensuring the stability and accuracy of the automated assembly process. The intelligent nut tightening control system includes the following functional modules:

[0072] Image acquisition module: Real-time acquisition of point cloud images of the working environment through laser galvanometer stereo camera 27.

[0073] Image Processing Module: This module uses the Open3D point cloud processing library for data preprocessing, point cloud feature extraction, and target recognition. It comprises several sub-modules, including a data preprocessing module, a feature extraction and matching module, a point cloud registration module, a result optimization and verification module, and a pose calculation module. The data preprocessing module employs operations such as filtering based on point cloud position distance, noise reduction, and outlier removal to improve point cloud data quality. The feature extraction and matching module uses techniques such as normal vector calculation and keypoint detection for initial estimation of point cloud registration. The point cloud registration module combines RANSAC and ICP algorithms to achieve accurate point cloud alignment. The result optimization and verification module performs error analysis on the registration results to optimize matching accuracy. The pose calculation module calculates the optimal grasping pose of the robotic arm using the SVD algorithm and transforms the visual coordinates to the robotic arm coordinate system.

[0074] Robotic arm control module: This module acquires displacement and attitude information in real time through sensors inside the robotic arm, enabling high-precision control. This includes path planning and real-time feedback. Path planning is based on Dynamic Movement Primitives (DMP) to learn the taught trajectory and optimize the path using obstacle avoidance algorithms. Inverse kinematics is used to calculate joint angles, generating efficient, collision-free motion trajectories. Specifically, such as... Figure 6 As shown, the robotic arm's movement path follows a diagonal installation sequence, causing the end effector 2 to tighten each nut on the working surface one by one, repeating this process twice to ensure the accuracy and stability of nut installation and prevent omissions or misinstallation. Real-time feedback is provided via image data acquired by a torque sensor and a laser galvanometer stereo camera 27 to dynamically adjust the robotic arm's movements. Preferably, the robotic arm control module includes limiting and alarm mechanisms, such as load changes and joint limits, to ensure safe operation of the robotic arm. The pre-defined robotic arm movement program achieves fully automated operation, reducing manual intervention, improving work efficiency, and lowering the operator's workload.

[0075] Torque control module: By monitoring torque changes in real time, it prevents over-tightening that could damage threads or cause incomplete assembly. For example... Figure 7 As shown, different locking parameters, such as locking angle and speed, are set according to different types of nuts and ball valve designs. A staged tightening strategy is adopted: a small torque for rapid tightening in the pre-tightening stage, and high-precision torque control in the final stage. Considering the special material and load requirements of the turbine ball valve nuts, the torque-angle method and yield point control method are integrated to further optimize the torque curve, ensuring the best tightening effect during assembly and ensuring the reliability of the nut connection. An electromagnetic torque sensor is used to monitor the mechanical changes during the tightening process, ensuring that each nut maintains stable pressure throughout the assembly process and avoiding over-tightening or over-loosening. Furthermore, a force feedback mechanism optimizes fine-tuning during the assembly process.

[0076] Dual Protection Module: During assembly, nuts may jam due to improper positioning or excessive torque, affecting production efficiency and potentially damaging equipment. To address this, the system program incorporates a dual protection module, including a force sensor monitoring module and a vision monitoring module. Force Sensor Monitoring Module: The force sensor on the end effector 2 detects the force applied to the nut in real time. When the friction or torque exceeds a set value, the system automatically stops and issues an alarm, prompting the operator to check. Vision Monitoring Module: The system's laser galvanometer stereo camera 27 monitors the assembly process in real time and feeds feedback to the system program, ensuring the correct position and orientation of the nut and preventing jamming. This dual monitoring mechanism significantly improves the stability and safety of the assembly process.

[0077] Data Management Module: Records assembly data for each ball valve, including the nut installation sequence and tightening torque, and maps this data to a unique identifier for each ball valve. Furthermore, the data management module supports cloud storage, enabling data access and analysis at any time, improving overall production transparency and traceability. The data management module includes a data recording and analysis module and a fault tolerance and recovery module. Data Recording and Analysis Module: Records relevant data during robot operation, such as nut position, installation sequence, tightening torque, speed changes, and tension changes, facilitating quality traceability and maintenance management. Fault Tolerance and Recovery Module: Considers various possible error scenarios, such as sensor failure and power outages, and has corresponding fault tolerance and recovery mechanisms.

[0078] An automatic assembly method for a large nut on a water turbine ball valve according to the present invention includes the following steps:

[0079] Step 1: Target detection and localization: such as Figure 8 As shown, the system guides the robotic arm 1 to perform precise assembly operations based on the target's position and posture output by the vision recognition mechanism. Simultaneously, the system displays the robotic arm's working status and assembly progress through a real-time monitoring interface, allowing operators to make adjustments and optimizations. In the assembly process of the turbine ball valve, precise positioning is crucial to ensuring the robotic arm's accurate operation. Therefore, the system employs advanced vision recognition technology, primarily using a laser galvanometer stereo camera 27. The laser galvanometer stereo camera 27 acquires the workpiece's three-dimensional point cloud data in real time by emitting a laser beam, combined with a high-precision image sensor. After feature extraction using image processing algorithms, the system can accurately identify the position of the nut and screw hole and calculate the robotic arm's motion trajectory. Specifically, the steps include the following:

[0080] Step 1.1: Image Acquisition and Data Preprocessing: A laser galvanometer stereo camera 27 is used to acquire 3D point cloud data and image information in real time to accurately obtain the spatial coordinates and geometry of the nut installation area. Voxel Grid filtering: Downsampling the original point cloud improves computational efficiency. Statistical Outlier Removal: Outliers are removed, optimizing the point cloud data quality. Radius Outlier Removal: Isolated points are further removed, improving the coherence of the point cloud.

[0081] A laser emits a high-precision laser beam, guided by a galvanometer system, to perform high-speed scanning on the surface of the object being measured. By controlling the deflection angle of the galvanometer, line scanning or area scanning modes can be achieved to adapt to different measurement needs. When the laser shines on the surface of the object, partial reflection occurs, and the reflected light is captured by a high-resolution image sensor in the camera. The sensor records data such as the position of the laser spot, reflection intensity, and time delay. The distance from the object surface to the camera is calculated by measuring the time from laser emission to return using the Time-of-Flight (ToF) method. Alternatively, triangulation can be used to calculate depth information based on the angular difference between the incident and reflected light. Combining multi-point measurements, complete 3D point cloud data is formed. The collected data is processed by algorithms to form high-precision point cloud data. Through filtering, registration, and reconstruction, a complete 3D model of the object is generated, which can be used for detection, measurement, or identification. The data is transmitted to an intelligent nut tightening control system via standard interfaces (such as USB, Ethernet, ROS protocol, etc.) for intelligent analysis and application.

[0082] During the measurement process, the galvanometer system quickly adjusts the direction of the laser beam to cover the entire nut mounting area, while the image sensor records the three-dimensional information of the area at a specified frame rate to ensure that the captured data is accurate and reliable.

[0083] Step 1.2: Image Processing and Feature Recognition: Image processing is the core part of nut installation point detection. The system uses a series of image processing techniques to identify the specific position and angle of the nut, including point cloud feature extraction, coarse registration, and fine registration.

[0084] Step 1.2.1: Point cloud feature extraction: Calculate the normal vector information of the point cloud to provide a basis for subsequent feature matching.

[0085] Step 1.2.2: Coarse registration: The local geometric features of the point cloud are extracted using FPFH (Fast Point Feature Histogram) to achieve preliminary point cloud matching; the initial transformation estimation is performed based on the Random Sample Consensus Algorithm (RANSAC) to obtain a relatively accurate rotation matrix and translation vector; the sampling consistency-based iterative registration algorithm SAC-IA is used to find the initial transformation matrix through feature matching, thereby improving the convergence speed of ICP.

[0086] Step 1.2.3: Fine registration: Iterative nearest point algorithm ICP is used: Based on the least squares optimization method, the point cloud is accurately registered; Weighted ICP: Based on the traditional ICP, the matching point pairs are assigned weights to improve the matching accuracy; Color ICP: RGB information is used to optimize the point cloud matching and improve the registration effect.

[0087] Step 1.3: Pose Calculation: Based on the nut's position and orientation identified by the vision system, the spatial pose of the nut is calculated to guide the robotic arm 1 in accurate grasping and assembly. The nut's installation point position information is transmitted to the intelligent nut tightening control system through the transformation relationship between the vision coordinate system and the robotic arm coordinate system.

[0088] 3D Reconstruction: Using calibrated camera parameters, the 2D image coordinates of the nut are converted into 3D spatial coordinates, and point cloud data of the nut's mounting points is generated. Based on the camera's intrinsic and extrinsic parameters, the following formula is used to convert the 2D image coordinates ( u , v Convert to three-dimensional spatial coordinates (X,Y,Z):

[0089]

[0090] in R For rotation matrix, T This is the translation vector. Through this transformation, the precise position of the nut in three-dimensional space can be obtained.

[0091] Pose determination: The optimal pose for the robotic arm to grasp the nut is calculated using the Singular Value Decomposition (SVD) algorithm, ensuring that the end effector 2 can correctly align with the nut. This process uses feature points from point cloud data to fit the optimal grasping posture of the nut.

[0092] Step 2: Path Planning and Execution: Combining the data recognized by the vision recognition mechanism, the system automatically generates the gripping and assembly paths for the robotic arm. The system instructs the robotic arm 1 to move along the optimal path to the nut installation point via a control program, ensuring efficient and accurate operation. The robotic arm 1 moves along a diagonal installation sequence, causing the end effector 3 to tighten each nut on the working surface one by one, repeating the operation twice to ensure the accuracy and stability of nut installation and prevent omissions or incorrect installations.

[0093] The DMP algorithm is used to optimize the path of the robotic arm, ensuring that the path does not interfere with other equipment or obstacles. The optimized path can effectively reduce movement time and improve assembly efficiency.

[0094] Step 3: Assembly execution and control: The robotic arm grabs the nut and positions it to the nut through the end effector 2, and performs the tightening action. The floating compensation mechanism compensates for the error, so that the end effector 2 has the ability to adapt during the grabbing, rotating and tightening process, avoiding assembly errors caused by rigid collisions; and improving assembly accuracy.

[0095] Staged tightening strategy: small torque for rapid tightening during the pre-tightening stage, and high-precision torque control in the final stage to ensure proper tightening.

[0096] Step 4: Quality Monitoring and Feedback: Real-time Feedback and Adjustment: During the robotic arm's path planning process, the system receives real-time sensor feedback data and adjusts the posture to ensure precise alignment between the end effector 2 and the nut mounting point. During nut tightening, the force sensor monitors the applied force on the nut in real time. If the torque exceeds the set threshold, the system immediately stops operation and issues an alarm, prompting the operator to perform a manual inspection to prevent the nut from jamming or being damaged.

[0097] Visual correction and position calibration: During the assembly process, the system continuously adjusts the posture of the robotic arm through visual feedback to correct deviations caused by nut position errors or angle deviations, ensuring that the nut is correctly assembled in place.

[0098] Foreign object detection in screw holes: The system uses infrared or ultrasonic sensors to monitor whether there are foreign objects inside the assembly holes. When an abnormal distance or an obstacle is detected, the system will immediately stop the assembly process and notify the operator.

[0099] In use, after the robotic arm 1 is installed at the designated working point on the platform, the intelligent nut tightening system is activated to initialize the robotic arm and tightening device. The laser galvanometer stereo camera 27 accurately locates the working area, identifies the bolt, the end effector 2 grasps the bolt, the laser galvanometer stereo camera 27 identifies the installation point, and the end effector 2 performs the tightening operation. The intelligent nut tightening control system controls and regulates the robotic arm 1 to accurately perform the installation action at each working point in a set sequence. After completing the work, the ball valve is flipped, and the robotic arm 1 is installed on another platform of the ball valve to continue the work.

[0100] It is understood that the present invention has been described through some embodiments, and those skilled in the art will recognize that various changes or equivalent substitutions can be made to these features and embodiments without departing from the spirit and scope of the invention. Furthermore, under the teachings of the present invention, these features and embodiments can be modified to adapt to specific situations and materials without departing from the spirit and scope of the invention. Therefore, the present invention is not limited to the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application are within the protection scope of the present invention.

Claims

1. An automatic assembly system for a large nut of a turbine ball valve, comprising a robotic arm (1), an end effector (2), and an intelligent nut tightening control system, wherein the robotic arm (1) is installed at a designated working point on the platform of the turbine ball valve and is controlled by the intelligent nut tightening control system; the end effector (2) is installed at the end of the robotic arm (1) and is used to perform identification, gripping, and tightening operations; the intelligent nut tightening control system is used to control the operation of the robotic arm (1) and the end effector (2), characterized in that, The end effector (2) includes an electromagnet adsorption mechanism, a servo-driven rotation mechanism, a floating compensation mechanism, and a vision recognition mechanism. The servo-driven rotation mechanism and the vision recognition mechanism are mounted on the end effector (2). The electromagnet adsorption mechanism is mounted on the servo-driven rotation mechanism via the floating compensation mechanism. The electromagnet adsorption mechanism uses electromagnet adsorption to pick up nuts of different specifications. The servo-driven rotation mechanism is used to tighten the nuts. The floating compensation mechanism is used to automatically adjust the posture of the electromagnet (25) by extension and retraction during the nut screwing process, compensating for the installation. The visual recognition mechanism is used to identify the position, geometric features, and three-dimensional coordinates of the assembly point of the turbine ball valve nut in real time. The servo-driven rotation mechanism includes a mounting base (21), a servo motor (22), and a gripper (23). One end of the mounting base (21) is fixedly connected to the end of the robotic arm (1), and the other end is equipped with the servo motor (22), the gripper (23), and the electromagnet adsorption mechanism. The servo motor (22) is fixedly installed on the upper end of the mounting base (21), and the gripper (23) is installed on the lower end of the mounting base (21). The gripper (23) includes a base. (231) and claws (232), the base (231) is fixedly connected to the output end of the servo motor (22), the claws (232) are distributed around the base (231) and fixedly connected to the base (231), a lever (24) is installed on the side of the nut to be installed, the servo motor (22) rotates to drive the gripper (23) to rotate, the claws (232) of the gripper (23) move the lever (24) to tighten the nut; the electromagnet adsorption mechanism includes an electromagnet (25), the electromagnet (25) is installed on the base of the gripper (23) through a floating compensation mechanism ( Below 231), the electromagnet (25) dynamically adjusts the magnetic force according to the material and size of the nut; the floating compensation mechanism includes multiple long screws (26), which are evenly distributed on the base (231) of the gripper (23). The long screws (26) are telescopically slidably connected to the base (231). The upper end of the long screw (26) has a nut (261) to limit the extension length of the long screw (26) and ensure that the long screw (26) does not come out of the base (231); the lower end of the long screw (26) is fixedly connected to the electromagnet (25).

2. The automatic assembly system for the large nut of the turbine ball valve according to claim 1, characterized in that, The visual recognition mechanism includes a laser galvanometer stereo camera (27), which is installed in the middle of the mounting base (21) and is used to identify the pose, geometric features of the nut, and three-dimensional coordinates of the assembly point of the turbine ball valve nut in real time. The laser galvanometer stereo camera (27) includes a laser, a galvanometer system, and an image sensor. The laser is responsible for emitting a high-precision laser beam. The galvanometer system includes a high-speed scanning galvanometer or an optical microelectromechanical system (MEMS) galvanometer, which is used to adjust the scanning direction of the laser beam. The image sensor uses a high-sensitivity CMOS or CCD sensor to calculate the flight time or angular deviation of the laser signal, thereby determining the depth information of the object surface.

3. The automatic assembly system for the large nut of the turbine ball valve according to claim 1, characterized in that, An electromagnetic torque sensor and a force sensor are installed on the end effector (2) to monitor the torque and mechanical changes during the locking process in real time.

4. The automatic assembly system for the large nut of the turbine ball valve according to claim 1, characterized in that, The intelligent nut tightening control system includes the following functional modules: Image acquisition module: Real-time acquisition of point cloud images in the working environment via laser galvanometer stereo camera (27); Image processing module: Utilizing the Open3D point cloud processing library for data preprocessing, point cloud feature extraction, and target recognition, the image processing module includes a data preprocessing module, a feature extraction and matching module, a point cloud registration module, a result optimization and verification module, and a pose calculation module. The data preprocessing module filters, reduces noise, and removes outliers based on point cloud position and distance. The feature extraction and matching module performs initial estimation for point cloud registration. The point cloud registration module combines RANSAC and ICP algorithms to achieve accurate point cloud alignment. The result optimization and verification module performs error analysis on the registration results. The pose calculation module calculates the optimal grasping pose of the robotic arm using the SVD algorithm and transforms the visual coordinates to the robotic arm coordinate system. Robotic arm control module: The robot arm acquires displacement and posture information in real time through sensors inside the robot arm and performs high-precision control, including path planning and real-time feedback. The path planning is based on the teaching trajectory learned by Dynamic Movement Primitives and the path is optimized by combining obstacle avoidance algorithm. The joint angle is calculated by inverse kinematics to generate an efficient collision-free motion trajectory. The movement path of the robot arm (1) is arranged in a diagonal installation order so that the end effector (2) tightens each nut on the working surface one by one and repeats the operation twice. The real-time feedback dynamically adjusts the robot arm's movements through image data acquired by force sensor and laser galvanometer stereo camera (27). Torque control module: Prevents overtightening by monitoring torque changes in real time; sets different locking parameters according to different types of nuts and ball valves. A phased tightening strategy is adopted: a small torque for rapid tightening in the pre-tightening stage, and high-precision torque control in the final stage; Integrating torque-angle method and yield point control method to optimize torque curve; using electromagnetic torque sensor to monitor mechanical changes during locking process to ensure that each nut maintains stable pressure throughout the assembly process; optimizing fine-tuning during assembly process through force feedback mechanism; Dual protection module: including force sensor monitoring module and vision monitoring module. The force sensor monitoring module includes a force sensor on the end effector (2) to detect the force on the nut in real time. When the friction or torque exceeds the set value, the system will automatically stop and issue an alarm to prompt the operator to check. The vision monitoring module includes the system's laser galvanometer stereo camera (27) to monitor the assembly process in real time and feed it back to the system program to ensure the correct position and orientation of the nut. Data Management Module: Records the assembly data of each ball valve and maps this data to the unique identifier of each ball valve. The data management module supports cloud storage. The data management module includes a data recording and analysis module and a fault tolerance and recovery module. The data recording and analysis module records relevant data during the robot's operation. The fault tolerance and recovery module considers various possible error situations and has corresponding fault tolerance and recovery mechanisms.

5. An automatic assembly method for an automatic assembly system of a large nut for a turbine ball valve as described in any one of claims 1-4, characterized in that, Includes the following steps: Step 1: Target detection and positioning: The system guides the robotic arm (1) to perform precise assembly operations based on the position and posture of the target output by the visual recognition mechanism. At the same time, the working status and assembly progress of the robotic arm are displayed through the real-time monitoring interface. The laser galvanometer stereo camera (27) emits a laser beam and, combined with a high-precision image sensor, acquires the three-dimensional point cloud data of the workpiece in real time. Step 2: Path planning and execution: Based on the data identified by the vision recognition mechanism, the gripping path and assembly path of the robotic arm are automatically generated. The system controls the robotic arm (1) to move to the nut installation point along the optimal path through the control program. The movement path of the robotic arm (1) follows the diagonal installation sequence, so that the end effector (2) tightens each nut on the working surface one by one, and repeats the operation twice. The DMP algorithm is used to optimize the path of the robotic arm to ensure that the path does not interfere with other equipment or obstacles. Step 3: Assembly execution and control: The robotic arm (1) grabs the nut and positions it to the nut through the end effector (2), and performs the tightening action, and compensates for the error through the floating compensation mechanism; Staged tightening strategy: small torque for rapid tightening in the pre-tightening stage, and high-precision torque control in the final stage to ensure proper tightening; Step 4: Quality monitoring and feedback: Real-time feedback and adjustment: During the path planning process of the robotic arm, the system will receive sensor feedback data in real time and adjust the posture to ensure the precise alignment of the end effector (2) and the nut mounting point; During the nut tightening process, the force sensor monitors the applied force on the nut in real time. If the torque exceeds the set threshold, the system will immediately stop the operation and issue an alarm to prompt the operator to perform a manual inspection to prevent the nut from jamming or being damaged.

6. The automatic assembly method according to claim 5, characterized in that, Step 1 includes the following steps: Step 1.1: Image acquisition and data preprocessing: Use a laser galvanometer stereo camera (27) to acquire three-dimensional point cloud data and image information in real time to accurately obtain the spatial coordinates and geometry of the nut installation area; Voxel filtering: downsample the original point cloud; Statistical filtering: remove outliers and optimize the point cloud data quality; Radius filtering: further remove isolated points; Step 1.2: Image Processing and Feature Recognition: Point cloud feature extraction, coarse registration, and fine registration are used to identify the specific position and angle of the nut; Step 1.3: Pose calculation: Based on the position and orientation of the nut identified by the vision system, calculate the spatial pose of the nut to guide the robotic arm (1) to accurately grasp and assemble it; through the transformation relationship between the vision coordinate system and the robotic arm coordinate system, transmit the nut installation point position information to the intelligent nut tightening control system; 3D Reconstruction: Using the calibrated camera parameters, the 2D image coordinates of the nut are converted into 3D spatial coordinates, and point cloud data of the nut's mounting points are generated. Based on the camera's intrinsic and extrinsic parameters, the 2D image coordinates (u,v) are converted into 3D spatial coordinates (X,Y,Z) using the following formula: Where R is the rotation matrix and T is the translation vector; through this transformation, the precise position of the nut in three-dimensional space is obtained; Pose determination: The optimal pose of the robotic arm when grasping the nut is calculated by using the singular value decomposition algorithm SVD. The best grasping posture of the nut is fitted using the feature points of the point cloud data. Visual correction and position calibration: During the assembly process, the system continuously adjusts the posture of the robotic arm through visual feedback to correct deviations caused by nut position errors or angle deviations, ensuring that the nut is correctly assembled in place; Foreign object detection in screw holes: Infrared or ultrasonic sensors are used to monitor whether there are foreign objects inside the assembly holes. When an abnormal distance or an obstacle is detected, the system immediately stops the assembly process and notifies the operator.

7. The automatic assembly method according to claim 6, characterized in that, Step 1.2 includes the following steps: Step 1.2.1: Point cloud feature extraction: Calculate the normal vector information of the point cloud to provide a basis for subsequent feature matching; Step 1.2.2: Coarse registration: Fast Point Feature Histogram (FPFH) is used to extract local geometric features of the point cloud to achieve preliminary point cloud matching; the initial transformation estimation is performed based on the Random Sample Consensus Algorithm (RANSAC) to obtain a relatively accurate rotation matrix and translation vector; the sampling consensus iterative registration algorithm (SAC-IA) is used to find the initial transformation matrix through feature matching. Step 1.2.3: Fine registration: Iterative nearest point algorithm ICP is used: Based on the least squares optimization method, the point cloud is accurately registered; Weighted ICP: Based on the traditional ICP, the matching point pairs are assigned weights to improve the matching accuracy; Color ICP: RGB information is used to optimize the point cloud matching.

Images