A bolt automatic locking path planning and control method and system

Abstract

The application discloses a bolt automatic locking path planning and control method and system, and relates to the technical field of precise assembly automation. The method comprises the following steps: S1: according to the identification result of the workpiece to be locked, the corresponding locking strategy is determined from the material-process parameter mapping table, and the corresponding feedforward compensation signal is obtained through the locking strategy and the rigid-flexible coupling dynamics model; S2: according to the feedforward compensation signal, the six-dimensional force sensor data, the locking strategy, the motor current size and the encoder data, the outer ring position and the inner ring force are regulated and controlled; S3: the actual data in the tightening process is obtained, the actual torque-angle relationship curve is constructed, and the corresponding locking quality is determined through the comparison result between the actual torque-angle relationship curve and the target torque-angle relationship curve. The application can ensure that the bolt can be accurately aligned with the target hole position in a complex motion trajectory, and effectively suppress the vibration and deformation under high acceleration or large range rotation working conditions.

Description

Technical Field

[0001] This invention relates to the field of precision assembly automation technology, specifically to a method and system for automatic bolt fastening path planning and control. Background Technology

[0002] In recent years, automated bolt fastening systems based on industrial robots have gradually become a research hotspot. By integrating electric or pneumatic tightening tools into the end effector of a multi-degree-of-freedom robotic arm, and combining them with vision recognition, force control feedback, and path planning algorithms, it is possible to automatically identify, locate, and fasten multiple bolt holes on complex workpieces.

[0003] Chinese invention patent application CN120762282A discloses an adaptive control method and system for robots in dynamic environments. Based on multi-sensor fusion and real-time anomaly detection, it combines impedance-compliant control to achieve dynamic compensation. Through lightweight error models and trajectory interpolation techniques, computational efficiency is improved compared to traditional model predictive control, supporting high-frequency real-time correction. The system adopts a modular design, adapting to heterogeneous robot platforms by adjusting equivalent mass-damping-spring parameters. For sudden disturbances, it integrates trajectory smoothing and impedance control techniques, reducing system oscillation amplitude and effectively suppressing execution jitter. This method achieves high-precision force tracking and stable control in scenarios such as walking in complex terrain and dynamic grasping, solving problems such as response lag, strong model dependence, and actuator overheating in traditional fixed-parameter control, significantly improving the robot's adaptability to dynamic environments and task completion rate.

[0004] However, the above and similar technical solutions still have the following shortcomings: most current bolt fastening control strategies are based on feedback control of the current state, which cannot predict the dynamic impact of the robot arm's kinematic characteristics, link flexibility and load changes on the end-positioning accuracy. Moreover, for workpieces of different materials (such as steel and aluminum), the same control parameters may lead to unstable fastening quality, which seriously restricts production accuracy and flexibility. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for automatic bolt fastening path planning and control, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for automatic bolt fastening path planning and control, comprising:

[0007] S1: Feedforward prediction compensation: Based on the identification result of the workpiece to be locked, the corresponding locking strategy is determined from the set material-process parameter mapping table, and the corresponding feedforward compensation signal is obtained through the locking strategy and the constructed rigid-flexible coupling dynamic model.

[0008] S2: Feedback Adaptive Control: Based on the feedforward compensation signal, six-dimensional force sensor data, locking strategy, motor current magnitude, and encoder data, the outer loop position and inner loop force are adjusted;

[0009] S3: Quantitative evaluation and processing: Based on the outer ring position and inner ring force of the control, the locking task is performed, and the corresponding actual data of the tightening process is obtained to construct the actual torque-angle relationship curve. At the same time, the corresponding locking quality is determined by comparing the actual torque-angle relationship curve with the target torque-angle relationship curve.

[0010] Furthermore, the corresponding feedforward compensation signal is obtained, including:

[0011] S1.1: Perception Mapping: The workpiece to be locked is identified through the interface of the set visual recognition system, RFID reading and writing system and manufacturing execution system, and the corresponding process data is determined according to the recognition result. At the same time, the material-process parameter mapping table is queried according to the process data to determine the corresponding locking strategy.

[0012] S1.2: Model Construction: Based on the kinematic structure of the robotic arm and the CAD drawings, set the physical parameters and topology of each kinematic structure, and construct a rigid body model based on the physical parameters and topology. At the same time, perform modal analysis on the kinematic structure to obtain the corresponding modal coordinates, and introduce the modal coordinates into the rigid body model to obtain the corresponding rigid-flexible coupling dynamic model.

[0013] S1.3: Real-time prediction: The current position, target hole position and motion trajectory of the workpiece to be locked are all used as inputs to the rigid-flexible coupling dynamic model. The corresponding predicted pose deviation is obtained as output. At the same time, the predicted pose deviation is inverted to obtain the corresponding feedforward compensation signal.

[0014] Furthermore, the rigid-flexible coupling dynamic model and the three-dimensional CAD model of the robotic arm are integrated to obtain an integrated robotic arm model. The integrated robotic arm model is then assembled using a digital twin platform to construct a digital twin model. Simultaneously, based on the feedforward compensation signal, simulation deduction is performed on the digital twin model to obtain the corresponding simulation prediction pose deviation.

[0015] Furthermore, the simulated predicted pose deviation is compared with the allowable pose deviation range in the locking strategy, and based on the comparison result, the final feedforward compensation signal is determined, specifically as follows:

[0016] When the simulated predicted pose deviation is within the allowable pose deviation range, the feedforward compensation signal is the final feedforward compensation signal; otherwise, the feedforward compensation signal is corrected until the simulated predicted pose deviation is within the allowable pose deviation range.

[0017] Furthermore, the feedforward compensation signal is corrected, including:

[0018] S1.4.1: Problem Analysis: Based on the simulation process of the digital twin model, the collision area in the simulation process is determined, and based on the distribution of pose deviation along the path in the collision area, the corresponding problem type is determined, including geometric interference problem and dynamic performance problem;

[0019] S1.4.2: Geometric processing: Adjust the joint angle values ​​of the robotic arm, set multiple similar posture values, and adjust the joint posture values ​​in the simulation process of the digital twin model according to the similar posture values. Perform local collision detection simulation on the simulation area where the similar posture values ​​are located. At the same time, determine the safe and collision-free similar posture values ​​according to the local collision detection simulation results, and re-simulate and verify the feedforward compensation signal according to the safe and collision-free similar posture values.

[0020] S1.4.3: Performance processing: Based on the problem section identified during the initial simulation of the digital twin model, the problem section is divided into multiple path points, and intermediate transition points are inserted into the path points. At the same time, the velocity curve and acceleration curve corresponding to the problem section are smoothed through a graphical curve editing algorithm.

[0021] Furthermore, the position of the outer ring and the force of the inner ring are controlled, including:

[0022] S2.1: Position Adjustment: The three-dimensional force and three-dimensional torque of the end effector are obtained by setting a six-dimensional force sensor. The target force and target torque are determined by setting an adaptive impedance control law. The corresponding pose correction command is obtained according to the force deviation and torque deviation between the three-dimensional force and the target force, and between the three-dimensional torque and the target torque. The pose correction command and the final feedforward compensation signal are superimposed to obtain the corresponding target correction command.

[0023] S2.2: Force Adjustment: Through the locking strategy, the corresponding optimal tightening curve is determined. The current magnitude and encoder data of the electric screwdriver's built-in motor are obtained through the set encoder and current sensor. Based on the current magnitude and encoder data, a real-time torque-angle relationship curve is constructed. At the same time, the real-time torque-angle relationship curve is compared with the optimal tightening curve. Based on the comparison result, the torque and angle during the tightening process are adjusted.

[0024] S2.3: Cooperative Operation: According to the target correction command, the outer loop controller moves to put the bolt tip into the hole, while the inner loop controller starts and rotates the screwdriver.

[0025] Furthermore, the corresponding locking quality is determined, including:

[0026] S3.1: Quality Quantification: Collect and acquire real-time torque and angle data of the electric screwdriver during the fastening process, construct the corresponding actual torque-angle relationship curve, and obtain the data difference corresponding to each time point in the torque-angle relationship curve based on the actual torque-angle relationship curve and the target torque-angle relationship curve, and determine the corresponding residual value based on the data difference.

[0027] S3.2: Quality Determination: The residual value is compared with a preset quality threshold range, and the corresponding locking quality is determined based on the comparison result, specifically as follows:

[0028] When the residual value is less than the lower limit of the preset quality threshold range, the corresponding locking quality is in the red zone and needs to be manually checked; when the residual value is within the preset quality threshold range, the corresponding locking quality is in the yellow zone and the parameters in the locking strategy are optimized; when the residual value is greater than the upper limit of the preset quality threshold range, the corresponding locking quality is in the green zone and the locking data is recorded.

[0029] Furthermore, it also includes:

[0030] S3.3: Strategy optimization: Combine the residual value and the gain coefficient to obtain the corresponding adjustment coefficient, and add the adjustment coefficient to the original target preload torque in the locking strategy to obtain the optimized target preload torque;

[0031] The slope of the original tightening rate curve in the locking strategy is adjusted to obtain the adjusted residual value between the adjusted tightening rate curve and the actual torque-angle relationship curve. At the same time, the adjusted residual value is compared with the original residual value to obtain the corresponding residual difference value. Based on the residual difference value, the optimized tightening rate curve is determined.

[0032] Furthermore, when the residual difference is less than 0, the corresponding adjusted tightening rate curve is the optimized tightening rate curve; otherwise, the slope of the original tightening rate curve is adjusted in the opposite direction according to the corresponding adjustment slope until the residual difference is less than 0.

[0033] An automatic bolt fastening path planning and control system uses any one of the above-mentioned automatic bolt fastening path planning and control methods.

[0034] Compared with the prior art, the beneficial effects of the present invention are:

[0035] Firstly, this invention, through the constructed rigid-flexible coupling dynamic model, can predict in advance the end-effector pose deviation caused by unmodeled dynamic factors such as flexible deformation of the robotic arm and load changes, and generate corresponding feedforward compensation signals. At the same time, through the simulation verification and correction mechanism of the digital twin model, the generated feedforward compensation signals can be further optimized, thereby ensuring that the bolt can still accurately align with the target hole position in complex motion trajectories. This effectively avoids positioning errors caused by response lag and effectively suppresses vibration and deformation under high acceleration or large-range rotation conditions.

[0036] Secondly, this invention can monitor the force state at the end in real time through a six-dimensional force sensor, and dynamically adjust the outer ring posture and inner ring tightening parameters through an adaptive impedance control algorithm, thereby ensuring that workpieces of different materials can obtain a stable axial locking force and avoiding over-tightening or stripping problems.

[0037] Thirdly, this invention generates the corresponding actual torque-angle relationship curve by using real-time torque-angle data after the locking is completed, and performs residual analysis with the target curve to quantify the corresponding locking quality. At the same time, through strategy optimization algorithm, the target pre-tightening torque or tightening rate curve is dynamically adjusted based on residual feedback, thereby forming a closed-loop iterative mechanism of perception-control-evaluation-optimization, which can continuously improve the stability and reliability of the locking process and reduce the cost of manual re-inspection.

[0038] Fourthly, this invention uses a digital twin model to simulate and extrapolate the feedforward compensation signal, which can identify potential geometric interference or dynamic performance problems during the movement of the robotic arm in advance. By adjusting joint posture, interpolating path points, and smoothing motion curves, the trajectory planning is actively optimized, thereby avoiding sudden collisions or vibrations during actual operation and improving the safety and reliability of the system. Attached Figure Description

[0039] Figure 1 This is a flowchart illustrating the automatic bolt fastening path planning and control method of the present invention;

[0040] Figure 2 This is a schematic diagram of the feedforward prediction compensation process in this invention;

[0041] Figure 3 This is a schematic diagram of the correction process for the feedforward compensation signal in this invention;

[0042] Figure 4This is a flowchart illustrating the feedback adaptive control method of the present invention;

[0043] Figure 5 This is a flowchart illustrating the quantitative evaluation processing method in this invention. Detailed Implementation

[0044] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0045] refer to Figure 1 This embodiment provides a method for automatic bolt fastening path planning and control, which specifically includes the following steps:

[0046] Step S1: Feedforward Prediction Compensation. This involves identifying the workpiece to be locked, obtaining its material type, thickness, and threaded hole orientation information. Simultaneously, based on the workpiece's material type, the corresponding locking strategy is determined from the established material-process parameter mapping table. Specifically, the locking strategy in this embodiment includes the target preload torque, tightening rate curve, contact point determination threshold, and allowable orientation deviation range for each degree of freedom.

[0047] Furthermore, based on the determined allowable pose deviation range of each degree of freedom, during the movement of the robotic arm end effector toward the target hole, the end effector pose deviation caused by the unmodeled dynamics of the system is predicted by the constructed rigid-flexible coupling dynamic model, and the corresponding feedforward compensation signal is obtained based on the predicted end effector pose deviation.

[0048] Furthermore, through a digital twin platform, models of the robotic arm, workpiece, and fixture are constructed. A rigid-flexible coupling dynamic model is then loaded and set into the digital twin platform to simulate and deduce the acquired feedforward compensation signal, thereby obtaining the corresponding simulated end-effector pose deviation. Simultaneously, the acquired simulated end-effector pose deviation is compared with the corresponding allowable pose deviation range, and based on the comparison results, the feedforward compensation signal is corrected to obtain the final feedforward compensation signal.

[0049] Step S2: Feedback Adaptive Control. This involves controlling both the outer ring position and the inner ring force during the bolt tightening stage. Specifically, based on the final feedforward compensation signal obtained in Step S1 and the real-time six-dimensional force sensor data, the motion trajectory of the outer ring is adjusted. In other words, the end-target pose is dynamically corrected to ensure the bolt remains in the target axial locking state.

[0050] Furthermore, based on the locking strategy obtained in step S1, and the real-time motor current and encoder data, the screwdriver speed and downward pressure are adjusted through adaptive impedance control to make the tightening process conform to the set target torque-angle relationship curve.

[0051] Step S3: Quantitative Evaluation Processing. After a single tightening task is completed, the actual tightening process data is collected, and the residual value between the collected actual tightening process data and the target torque-angle relationship curve is obtained. The corresponding tightening quality is determined by comparing the obtained residual value with a preset quality threshold.

[0052] Furthermore, based on the determined locking quality, the target preload torque and tightening rate curves corresponding to the locking strategy obtained in step S1 are optimized using a set optimization algorithm, such as a rule adjustment algorithm.

[0053] This embodiment also provides an automatic bolt fastening path planning and control system, which uses the above-mentioned automatic bolt fastening path planning and control method.

[0054] In this embodiment, by constructing a rigid-flexible coupling dynamic model and determining the allowable pose deviation range for each degree of freedom, the corresponding end-effector pose deviation is predicted, and a corresponding feedforward compensation signal is generated. Simultaneously, a digital twin platform is used to simulate and extrapolate the generated feedforward compensation signal, obtaining the simulated end-effector pose deviation. Based on the obtained simulated end-effector pose deviation, the feedforward compensation signal is corrected. (Reference) Figure 2 This embodiment provides a feedforward prediction compensation method, which specifically includes the following steps:

[0055] Step S1.1: Perception Mapping. This involves identifying key information about the workpiece to be locked through the interfaces of the set vision recognition system, RFID reading and writing system, and manufacturing execution system. This includes material type, plate thickness, and the position and orientation of the threaded holes in space.

[0056] Specifically, in this embodiment, the visual recognition system places industrial cameras at key locations (such as the top or side) of the robotic arm's working unit and calibrates the industrial cameras. That is, the industrial cameras capture images of the workpieces to be locked, conveyed by the conveyor belt or positioning fixture, and the captured workpiece images are processed by a set deep learning model (such as an object detection network model or a segmentation network model) to identify the corresponding workpiece type and the image coordinates of the threaded hole center. Simultaneously, through coordinate transformation, the image coordinates of the threaded hole center are converted into the three-dimensional spatial pose of the robotic arm.

[0057] Furthermore, in this embodiment, the RFID reading and writing system affixes RFID tags to the trays or fixtures of the workpieces to be locked, and an RFID reader is installed at the entrance of the workstation. That is, the RFID reader reads the information from the RFID tags, including but not limited to the inherent properties of the workpiece, such as its material type and thickness.

[0058] Furthermore, the industrial control computer or robot controller interacts with the manufacturing execution system (MES) via industrial communication protocols (such as OPC UA and TCP / IP). In other words, the vision recognition system and RFID reader / writer system transmit the acquired process data to the MES through industrial communication protocols. It is worth noting that the process data stored in the MES in this embodiment also includes, but is not limited to, bolt specifications and required preload levels.

[0059] In this embodiment, the manufacturing execution system acquires the process data of the workpiece to be fastened, including but not limited to material type, thickness, and thread hole orientation information. Based on the acquired process data, the system queries a set material-process parameter mapping table to determine the corresponding fastening strategy. Specifically, the material-process parameter mapping table in this embodiment includes, but is not limited to, material type, thickness range, target torque, tightening speed curve ID, contact point force threshold, and allowable orientation deviation range.

[0060] Step S1.2: Model Construction. Based on the kinematic structure of the robotic arm and the corresponding CAD drawings, the physical parameters of each kinematic structure are determined, including length, center of mass position, and mass in a rigid model. Simultaneously, based on the determined topology of the kinematic structure, a corresponding dynamic modeling algorithm is set, such as the recursive Newton-Euler algorithm, to obtain the corresponding rigid body model.

[0061] Furthermore, the CAD model of each kinematic structure is imported into finite element analysis software (such as ANSYS or Abaqus) for modal analysis to determine the mode shapes and natural frequencies of each kinematic structure. Simultaneously, based on the degree of influence on dynamics, low-cost modes of a predetermined order are identified. These low-cost modes are then transformed using modal synthesis methods, such as the Craig-Bampton method, to obtain their corresponding modal coordinates. These modal coordinates are then embedded into the rigid body model to construct the corresponding rigid-flexible coupled dynamic model.

[0062] Step S1.3: Real-time prediction. Based on the current position of the workpiece to be locked and the corresponding target hole position, the path planning module determines the corresponding motion trajectory, including the theoretical curves of the angles, angular velocities, and angular accelerations of each joint of the robotic arm over time, to obtain the corresponding planned trajectory. Simultaneously, encoders on each joint obtain the angle value of each joint motor at the current moment, determining the actual initial state of the robotic arm, i.e., the current state. Based on the process data obtained in step S1.1, the corresponding current tool and workpiece information to be locked are determined, such as the electric screwdriver model and workpiece type. Based on the determined current tool and workpiece information, the corresponding load mass and inertia are determined to obtain the corresponding load information.

[0063] Furthermore, the obtained planned trajectory, current state, and load information are used as inputs to the rigid-flexible coupled dynamic model constructed in step S1.2, and the corresponding predicted pose deviations are output, including the predicted position deviations and predicted attitude angle deviations on each directional axis.

[0064] Furthermore, by using an inversion algorithm, the obtained predicted pose deviation is inverted to obtain the corresponding reverse predicted pose deviation. In other words, the obtained reverse predicted pose deviation is the corresponding feedforward compensation signal.

[0065] Step S1.4: Simulation Verification. This involves integrating the rigid-flexible coupling dynamic model obtained in Step S1.2 with the 3D CAD model of the robotic arm to obtain the corresponding integrated robotic arm model. Simultaneously, based on the actual layout of the physical workshop, the integrated robotic arm model, the end effector and tool model, and the workpiece and fixture model are assembled in a digital twin platform (such as NVIDIA Omniverse) to obtain the corresponding digital twin model.

[0066] Furthermore, based on the planned trajectory, current state, load information, and feedforward compensation signal obtained in step S1.3, simulation is performed in the constructed digital twin model to obtain the corresponding simulated predicted pose deviation. Simultaneously, the obtained simulated predicted pose deviation is compared with the allowable pose deviation range obtained in step S1.1 to correct the feedforward compensation signal, specifically as follows:

[0067] When the obtained simulated predicted pose deviation is within the allowable pose deviation range, the feedforward compensation signal obtained in step S1.3 does not need to be corrected and is the final feedforward compensation signal. Conversely, when the obtained simulated predicted pose deviation is not within the allowable pose deviation range, the feedforward compensation signal obtained in step S1.3 needs to be corrected. Then, based on the corrected feedforward compensation signal, step S1.4 is repeated until the obtained simulated predicted pose deviation is within the allowable pose deviation range.

[0068] In this embodiment, reference Figure 3 The obtained feedforward compensation signal is then corrected, specifically including the following steps:

[0069] Step S1.4.1: Problem Analysis. Based on the simulation process of the digital twin model, the collision areas in the simulation process are identified, and the distribution of pose deviations along the path is displayed graphically (e.g., color cloud maps and deviation vector arrows) at the collision areas to determine the corresponding problem type, including geometric interference problems and dynamic performance problems.

[0070] Specifically, the geometric interference problem in this embodiment includes purely geometric collisions caused by insufficient spatial clearance between the robotic arm, tool, or cable and the workpiece / fixture. The dynamic performance problem in this embodiment includes inaccurate end-effector positioning due to excessive flexible deformation of the robotic arm under specific motion states (such as high-speed start-stop, large-range rotation).

[0071] Step S1.4.2: Geometric Processing. When the identified problem type is a geometric interference problem, the joint angle values ​​of the robotic arm are adjusted in the digital twin model to set multiple similar posture values. Specifically, the similar posture values ​​in this embodiment can be adjusted based on the joint posture values ​​during the initial simulation and the preset posture adjustment values. It is worth noting that the preset posture adjustment values ​​in this embodiment can be specifically set according to actual needs, therefore, they are not specifically described in this embodiment.

[0072] Furthermore, based on the set similar attitude values, the corresponding simulation region is determined, and local collision detection simulation is performed within that region to determine each set similar attitude value, from which a safe and collision-free similar attitude value is identified. Simultaneously, based on the determined safe and collision-free similar attitude values, the digital twin model is reset, and the acquired feedforward compensation signal is re-simulated and verified.

[0073] Step S1.4.3: Performance Processing. When the identified problem type is a dynamic performance problem, the planned trajectory set during the simulation exercise is adjusted and optimized to reduce dynamic deformation. Specifically, based on the problem segment identified in the initial simulation, the problem segment is divided into multiple path points to construct the corresponding problem segment. Further, a numerical interpolation algorithm is used to insert multiple intermediate transition points into the problem segment to optimize the divided path points, thereby increasing the description of the problem segment, preventing the robotic arm from being in a high-acceleration state for extended periods, and thus reducing continuous deformation caused by inertial forces.

[0074] Furthermore, based on the velocity and acceleration curves corresponding to the identified problem sections, a graphical curve editing algorithm is used to smooth the velocity and acceleration curves, thereby reducing the corresponding maximum acceleration and deceleration. At the same time, an S-shaped (such as a square sine or polynomial) acceleration and deceleration pattern is set to avoid sudden impacts and the resulting strong vibrations.

[0075] In this embodiment, based on the acquired locking strategy and final feedforward compensation signal, and combined with the acquired motor current magnitude, encoder data, and six-dimensional force sensor data, the operation of the outer loop position and inner loop force are controlled respectively. (Reference) Figure 4 This embodiment provides a feedback adaptive control method, which specifically includes the following steps:

[0076] Step S2.1: Position Adjustment. This involves using a six-dimensional force sensor to acquire the three-dimensional force and torque of the end effector in real time, including radial force, axial force, and torque. Specifically, based on the adaptive impedance control law set in the outer loop controller, the force deviation and torque deviation between the three-dimensional force and the target force, and between the three-dimensional torque and the target torque, are determined by acquiring the three-dimensional force and torque in real time. It is worth noting that in this embodiment, both the target force and the target torque are set to zero.

[0077] Furthermore, based on the determined magnitude and direction of the force and torque deviations, the corresponding pose correction commands are obtained. Simultaneously, the final feedforward compensation signal obtained in step S1.4 is superimposed with the pose correction commands to obtain the corresponding target correction commands.

[0078] Step S2.2: Force Adjustment. Based on the locking strategy determined in Step S1.1, the optimal tightening curve corresponding to the workpiece to be locked is determined, and the operation of the inner loop controller is controlled according to the determined optimal tightening curve. Specifically, through the set encoder and current sensor, the current magnitude of the electric screwdriver's built-in motor and encoder data are acquired in real time, and the corresponding output torque and rotation angle are determined based on the acquired current magnitude and encoder data. In other words, the corresponding real-time torque-rotation angle relationship curve is obtained by determining the output torque and rotation angle.

[0079] Furthermore, the obtained real-time torque-angle relationship curve is compared with the set target torque-angle relationship curve (i.e., the optimal tightening curve) to ensure that the real-time torque-angle relationship curve closely matches the target torque-angle relationship curve. Specifically, when the real-time torque is less than the target torque, the screwdriver's downward pressure and angle are increased. When the real-time torque is greater than the target torque, the stiffness or damping is increased to prevent overshoot.

[0080] Step S2.3: Cooperative Operation. Based on the target correction command obtained in step S2.1, the robotic arm, under the control of the outer loop controller, inserts the bolt tip into the hole. Simultaneously, the inner loop controller starts and rotates the screwdriver.

[0081] Furthermore, when the bolt head contacts the surface of the workpiece to be fastened, the inner loop controller initiates the tightening phase, and adjusts the torque and rotation angle in real time during this phase. Simultaneously, as the preload increases, the outer loop controller adjusts the support force of the robotic arm in real time to maintain axial alignment. Once the target torque or rotation angle is reached, the inner loop controller stops operating, and the outer loop controller controls the robotic arm to move away.

[0082] In this embodiment, the locking quality is determined based on the residual value between the actual torque-angle relationship curve and the target torque-angle relationship curve after each locking task is completed. The locking strategy is then optimized based on the determined locking quality. (Reference) Figure 5 This embodiment provides a quantitative evaluation processing method, which specifically includes the following steps:

[0083] Step S3.1: Quality Quantification. During the tightening task, real-time torque and angle data of the electric screwdriver are collected, and a corresponding actual torque-angle relationship curve is constructed based on this data. Simultaneously, the constructed actual torque-angle relationship curve and the target torque-angle relationship curve are plotted on the same coordinate system, and the data difference between the actual torque-angle relationship curve and the target torque-angle relationship curve at each time point is obtained within the same coordinate system. Further, based on the data difference at each time point, the corresponding root mean square error is obtained, which is the magnitude of the corresponding residual value.

[0084] Step S3.2: Quality Determination. This involves comparing the residual value obtained in step S3.1 with a preset quality threshold range (it's worth noting that the preset quality threshold in this example can be set based on the average value of historical residual values ​​within a preset time period; that is, it can be specifically set according to actual data requirements, so it is not specifically described in this embodiment. Specifically, the upper limit of the preset quality threshold range is set to 5% of the average value, and the lower limit is set to 2% of the average value). Based on the comparison result, the corresponding locking quality is determined, specifically as follows:

[0085] When the obtained residual value is less than the lower limit of the preset quality threshold range, the corresponding locking quality is in the red zone and requires manual inspection. When the obtained residual value is within the preset quality threshold range, the corresponding locking quality is in the yellow zone, and the parameters in the locking strategy are optimized. When the obtained residual value is greater than the upper limit of the preset quality threshold range, the corresponding locking quality is in the green zone, and the locking data is recorded.

[0086] Step S3.3: Strategy Optimization. This involves optimizing the target preload torque and tightening rate curves in the locking strategy based on the residual values ​​obtained in step S3.1. Specifically, the obtained residual values ​​are combined with the set gain coefficient to obtain the corresponding adjustment coefficient. This adjustment coefficient is then combined with the original target preload torque in the locking strategy to obtain the optimized target preload torque. It should be noted that the gain coefficient set in this embodiment can be specifically set according to actual data requirements; therefore, it is not specifically described in this embodiment.

[0087] Furthermore, the slope of the original tightening rate curve in the locking strategy is adjusted, and the rate of decrease can be gradually reduced or increased according to actual needs. Simultaneously, the residual value between each adjusted tightening rate curve (i.e., the adjusted torque-angle relationship curve) and the actual torque-angle relationship curve is obtained; this is the adjusted residual value.

[0088] Furthermore, the adjusted residual value is compared with the residual value obtained in step S3.1 to obtain the corresponding residual difference. Simultaneously, based on the magnitude of the residual difference, the corresponding optimized torque-angle relationship curve is determined. Specifically, when the residual difference is less than 0, the corresponding adjusted torque-angle relationship curve is the optimized torque-angle relationship curve. Conversely, when the residual difference is not less than 0, the slope of the original tightening rate curve is adjusted in the opposite direction according to the corresponding adjustment slope until the residual difference is less than 0.

[0089] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended embodiments and their equivalents.

Claims

1. A method for automatic bolt fastening path planning and control, characterized in that, Including: S1: Feedforward Prediction Compensation: Based on the identification result of the workpiece to be locked, the corresponding locking strategy is determined from the set material-process parameter mapping table, and the corresponding feedforward compensation signal is obtained through the locking strategy and the constructed rigid-flexible coupling dynamic model, including: S1.1: Perception Mapping: The workpiece to be locked is identified through the interface of the set visual recognition system, RFID reading and writing system and manufacturing execution system, and the corresponding process data is determined according to the identification result. At the same time, the material-process parameter mapping table is queried according to the process data to determine the corresponding locking strategy. S1.2: Model Construction: Based on the kinematic structure and CAD drawings of the robotic arm, the physical parameters and topology of each kinematic structure are set. A rigid body model is constructed based on these physical parameters and topology. Modal analysis is performed on the kinematic structures to obtain the corresponding modal coordinates. These modal coordinates are then introduced into the rigid body model to obtain the corresponding rigid-flexible coupling dynamic model. The rigid-flexible coupling dynamic model and the 3D CAD model of the robotic arm are integrated to obtain an integrated robotic arm model. This integrated model is then assembled using a digital twin platform to construct a digital twin model. Simultaneously, simulations are performed on the digital twin model based on the feedforward compensation signal to obtain the corresponding simulated predicted pose deviation. The simulated predicted pose deviation is compared with the allowable pose deviation range in the locking strategy. Based on the comparison results, the final feedforward compensation signal is determined, specifically: When the simulated predicted pose deviation is within the allowable pose deviation range, the feedforward compensation signal is the final feedforward compensation signal; otherwise, the feedforward compensation signal is corrected until the simulated predicted pose deviation is within the allowable pose deviation range. S1.3: Real-time prediction: The current position, target hole position and motion trajectory of the workpiece to be locked are all used as inputs to the rigid-flexible coupling dynamic model. The corresponding predicted pose deviation is obtained as output. At the same time, the predicted pose deviation is inverted to obtain the corresponding feedforward compensation signal. The correction of the feedforward compensation signal includes: S1.4.1: Problem Analysis: Based on the simulation process of the digital twin model, the collision area in the simulation process is determined, and based on the distribution of pose deviation along the path in the collision area, the corresponding problem type is determined, including geometric interference problem and dynamic performance problem; S1.4.2: Geometric processing: Adjust the joint angle values ​​of the robotic arm, set multiple similar posture values, and adjust the joint posture values ​​in the simulation process of the digital twin model according to the similar posture values. Perform local collision detection simulation on the simulation area where the similar posture values ​​are located. At the same time, determine the safe and collision-free similar posture values ​​according to the local collision detection simulation results, and re-simulate and verify the feedforward compensation signal according to the safe and collision-free similar posture values. S1.4.3: Performance processing: Based on the problem section identified during the initial simulation of the digital twin model, the problem section is divided into multiple path points, and intermediate transition points are inserted into the path points. At the same time, the velocity curve and acceleration curve corresponding to the problem section are smoothed through a graphical curve editing algorithm. S2: Feedback Adaptive Control: Based on the feedforward compensation signal, six-dimensional force sensor data, locking strategy, motor current magnitude, and encoder data, the outer loop position and inner loop force are adjusted; S3: Quantitative evaluation and processing: Based on the outer ring position and inner ring force of the control, the locking task is performed, and the corresponding actual data of the tightening process is obtained to construct the actual torque-angle relationship curve. At the same time, the corresponding locking quality is determined by comparing the actual torque-angle relationship curve with the target torque-angle relationship curve.

2. The method for automatic bolt fastening path planning and control according to claim 1, characterized in that, The control of the outer ring position and the inner ring force includes: S2.1: Position Adjustment: The three-dimensional force and three-dimensional torque of the end effector are obtained by setting a six-dimensional force sensor. The target force and target torque are determined by setting an adaptive impedance control law. The corresponding pose correction command is obtained according to the force deviation and torque deviation between the three-dimensional force and the target force, and between the three-dimensional torque and the target torque. The pose correction command and the final feedforward compensation signal are superimposed to obtain the corresponding target correction command. S2.2: Force Adjustment: Through the locking strategy, the corresponding optimal tightening curve is determined. The current magnitude and encoder data of the electric screwdriver's built-in motor are obtained through the set encoder and current sensor. Based on the current magnitude and encoder data, a real-time torque-angle relationship curve is constructed. At the same time, the real-time torque-angle relationship curve is compared with the optimal tightening curve. Based on the comparison result, the torque and angle during the tightening process are adjusted. S2.3: Cooperative Operation: According to the target correction command, the outer loop controller moves to put the bolt tip into the hole, while the inner loop controller starts and rotates the screwdriver.

3. The method for automatic bolt fastening path planning and control according to claim 1, characterized in that, Determine the corresponding locking quality, including: S3.1: Quality Quantification: Collect and acquire real-time torque and angle data of the electric screwdriver during the fastening process, construct the corresponding actual torque-angle relationship curve, and obtain the data difference corresponding to each time point in the torque-angle relationship curve based on the actual torque-angle relationship curve and the target torque-angle relationship curve, and determine the corresponding residual value based on the data difference. S3.2: Quality Determination: The residual value is compared with a preset quality threshold range, and the corresponding locking quality is determined based on the comparison result, specifically as follows: When the residual value is less than the lower limit of the preset quality threshold range, the corresponding locking quality is in the red zone and needs to be manually checked; when the residual value is within the preset quality threshold range, the corresponding locking quality is in the yellow zone and the parameters in the locking strategy are optimized; when the residual value is greater than the upper limit of the preset quality threshold range, the corresponding locking quality is in the green zone and the locking data is recorded.

4. The method for automatic bolt fastening path planning and control according to claim 3, characterized in that, It also includes: S3.3: Strategy optimization: Combine the residual value and the gain coefficient to obtain the corresponding adjustment coefficient, and add the adjustment coefficient to the original target preload torque in the locking strategy to obtain the optimized target preload torque; The slope of the original tightening rate curve in the locking strategy is adjusted to obtain the adjusted residual value between the adjusted tightening rate curve and the actual torque-angle relationship curve. At the same time, the adjusted residual value is compared with the original residual value to obtain the corresponding residual difference value. Based on the residual difference value, the optimized tightening rate curve is determined.

5. The method for automatic bolt fastening path planning and control according to claim 4, characterized in that, When the residual difference is less than 0, the corresponding adjusted tightening rate curve is the optimized tightening rate curve; otherwise, the slope of the original tightening rate curve is adjusted in the opposite direction according to the corresponding adjustment slope until the residual difference is less than 0.

6. A bolt automatic fastening path planning and control system, characterized in that, The method for automatic bolt fastening path planning and control described in any one of claims 1-5 was used.

Images