Robot screwing method and system based on dynamic screw field and compliant control

Abstract

The application discloses a robot screwing method and system based on a dynamic spiral field and a compliant control, and the system comprises a perception module, a dynamic spiral field generator and a compliant coupling controller. The perception module calculates parameters such as screwing-in progress, radial deviation and meshing confidence in real time through multi-sensor fusion and a state observer, and estimates a friction coefficient and a contact state. The dynamic spiral field generator generates an intelligent screwing-in trajectory according to the perception state, and adaptively adjusts a pitch, a spiral radius and expected force and speed. The compliant coupling controller adopts parameter-adaptive axial lead control and rotary impedance control, and realizes high-precision and high-compliant force-position collaborative operation. The system can dynamically adapt to different screwing-in stages and abnormal working conditions, and improves the reliability, adaptability and operation precision of the screwing process.

Description

Technical Field

[0001] This invention belongs to the field of robotics, specifically relating to a method and system for screw tightening by a robot based on dynamic helical field and compliant control. Background Technology

[0002] Traditional robotic screw-tightening methods often rely on fixed trajectories and threshold judgments, lacking deep perception and dynamic adaptation capabilities to the screwing state. This frequently leads to problems such as alignment deviations, thread stripping, overtightening, or abnormal responses. In existing technologies, force control and position control often operate independently, failing to achieve intelligent coordination between the axial and rotational directions. Furthermore, they lack robustness under uncertainties such as changes in thread friction and hole position offset, limiting their application in demanding scenarios like precision assembly and aerospace. While existing robotic screw-tightening control methods introduce the concepts of force-position hybrid or impedance / admittance control in their control strategies, their core control parameters (such as stiffness and damping in the axial and rotational directions, and the pitch and radius of the expected trajectory) typically use fixed values ​​or preset values ​​based on simple stage switching. Adjustment strategies often remain at a qualitative or semi-quantitative level, failing to achieve continuous and smooth adaptation during a single screwing process, and further failing to coordinate the dynamic coupling between axial propulsion and rotational tightening. Summary of the Invention

[0003] This invention discloses a robotic screw-tightening system based on dynamic helical field and compliance control, comprising:

[0004] The sensing module is used to read sensor data, including axial force F. z Calculate the screw-in progress φ using torque τ, rotation angle θ, angular velocity ω, axial position z, and velocity v. prog Injection energy ratio ξ, radial deviation e radial Contact normal angle Φ contact The system calculates the engagement confidence level c and estimates the friction coefficient μ of the thread contact surface in real time using a state observer. Based on the friction coefficient μ and engagement confidence level c, the system classifies the current contact state.

[0005] A dynamic spiral field generator, connected to the sensing module, is used to dynamically adjust the adaptive pitch h and adaptive spiral radius R of the virtual spiral field based on the state parameters output by the sensing module, and to generate the desired axial force F. zd Desired axial trajectory z d and desired angular velocity ω d ;

[0006] A compliant coupling controller, connected to a dynamic helical field generator and a sensing module, is used to receive the desired axial force F. zd Desired axial trajectory z d and desired angular velocity ω d And combined with the real-time feedback of axial force Fz Given the position z, an axial acceleration command is calculated using a parameter-adaptive axial admittance control law; simultaneously, a rotational torque command is calculated using a parameter-adaptive rotational control law, which is either a feedforward-feedback composite control law or a rotational impedance control law.

[0007] Specifically, in the sensing module, the formula for calculating the spin-in energy ratio is: ξ(t) = (ξ raw (t)-ξ nom,adj (t)) / ξ nom,adj (t), ξ raw (t)=(τ(t)·ω(t)) / (F z (t)·v(t)+ε),

[0008] , where ξ nom ε is the nominal inclination energy ratio, κ is the numerical stability term, and κ is the nominal inclination energy ratio. fric μ is the friction correction coefficient. nom To design the coefficient of friction.

[0009] Specifically, in the perception module, the formula for calculating engagement confidence is:

[0010] Where N is the length of the sliding window, F z ' is the axial force F within the window z The mean of , θ' is the mean of the rotation angle θ within the window, β is the gain coefficient, ρ thresh The threshold value is used.

[0011] Specifically, the state observer in the sensing module is used to estimate the static friction coefficient μ. s and the coefficient of kinetic friction μ d Its update law is based on the measured torque, axial force, screw geometry parameters, and the difference between the theoretically predicted torque and the actual torque. The state observer outputs the comprehensive friction coefficient μ.

[0012] Specifically, the perception module uses a fuzzy logic classifier, taking the estimated friction coefficient μ and engagement confidence c as input, to classify the contact state into four modes: free engagement, initial engagement, normal engagement, and abnormal friction.

[0013] Specifically, in the dynamic helical field generator, the formula for calculating the adaptive pitch is:

[0014] Where h0 is the nominal pitch of the screw, α φ ,β φ ,α ξ ,β ξ This is for adjusting the coefficient.

[0015] Specifically, in the dynamic spiral field generator, the formula for calculating the adaptive spiral radius R includes a correction term based on radial deviation, and the correction gain decreases as the spiraling progress progresses.

[0016] Specifically, the axial admittance control law in the compliant coupling controller is: , of which M z e represents the axial virtual inertia. z For axial position error, stiffness K z and damping B z According to the screwing progress φ prog The spin-in energy ratio ξ and engagement confidence c are adaptively adjusted.

[0017] Specifically, the rotational control law in the compliant coupling controller adopts a rotational impedance control law:

[0018] , of which M θ It is the virtual inertia in the direction of rotation, B θ and K θ It is time-varying damping and stiffness, τ ext It is the external torque, stiffness K θ According to the screwing progress φ prog Adaptive adjustment of the engagement energy ratio ξ and meshing confidence c; coupling torque term τ couple Related to the axial force error, its coupling gain η is adaptively adjusted based on the spin-in energy ratio ξ and the friction coefficient μ.

[0019] This invention also discloses a working method for a robot screw-tightening system based on dynamic helical field and compliance control, comprising:

[0020] Sensing: Reading sensor data, including axial force F z Calculate the screw-in progress φ using torque τ, rotation angle θ, angular velocity ω, axial position z, and velocity v. prog Injection energy ratio ξ, radial deviation e radial Contact normal angle Φ contact The system calculates the engagement confidence level c and estimates the friction coefficient μ of the thread contact surface in real time using a state observer. Based on the friction coefficient μ and engagement confidence level c, the system classifies the current contact state.

[0021] Dynamic spiral field generation: Based on the state parameters output by the sensing module, the adaptive pitch h and adaptive spiral radius R of the virtual spiral field are dynamically adjusted, and the desired axial force F is generated. zd Desired axial trajectory z d and desired angular velocity ω d ;

[0022] Compliant coupling control: receiving the desired axial force F zd Desired axial trajectory zd and desired angular velocity ω d And combined with the real-time feedback of axial force F z Given the position z, an axial acceleration command is calculated using a parameter-adaptive axial admittance control law; simultaneously, a rotational torque command is calculated using a parameter-adaptive rotational control law, which is either a feedforward-feedback composite control law or a rotational impedance control law.

[0023] Beneficial technical effects: Deep state perception, through multi-sensor fusion and state observers, real-time estimation of key internal states such as friction coefficient, contact normal, and engagement confidence, achieving comprehensive understanding of the screwing process; Dynamic trajectory generation, based on real-time state adaptive adjustment of helical field parameters (pitch, radius, desired force / velocity), achieving smooth and adaptive screwing path planning, improving alignment accuracy and operational efficiency; Intelligent coordinated control, employing a compliant coupling strategy combining axial admittance control and rotational impedance control, achieving dynamic coordination of axial and rotational directions through an adaptive coupling mechanism, mimicking the "feel" of manual operation; Adaptable to different industry needs through parameter configuration, such as the high reliability requirements of aerospace, efficient batch assembly in automobile manufacturing, and precision operation of micro screws in electronic products. Attached Figure Description

[0024] Appendix Figure 1 This is a schematic diagram of a robot screw-tightening system based on dynamic helical field and compliant control according to the present invention.

[0025] Appendix Figure 2 This is a schematic diagram of a tilted screw. Detailed Implementation

[0026] This invention discloses a robotic screw-tightening method and system based on dynamic helical field and compliant control. The method, based on dynamic helical field and dual compliant control, solves technical problems in traditional screw-in systems such as alignment difficulties, excessive tightening and stripping, and abnormal response through real-time state perception and adaptive parameter adjustment. The system is used to implement this method. Figure 1 As shown, the system includes:

[0027] 1. Sensing Module: Reads sensor data, calculates state parameters, and updates internal state estimates by running the state observer, including friction coefficient, contact normal, and state mode classification.

[0028] During screw insertion, the robot needs a comprehensive understanding of the current insertion state, including directly measurable parameters (such as position and force) and internal states that are not directly measurable but are crucial for control (such as the coefficient of friction and the condition of the contact surface). Traditional methods rely solely on simple threshold judgments and lack a deep understanding of the insertion state, resulting in an inability to intelligently handle complex situations.

[0029] The sensing module of this invention establishes a multi-layered state sensing system, capable of calculating direct state parameters in real time and estimating internal states through observers. This gives the system a "feel" similar to that of a human craftsman, enabling it to sense key information such as whether the screwing is smooth, whether it is aligned, and whether the thread engagement is good. The direct measurement sensor system includes a six-dimensional force / torque sensor, which measures force F in three directions. x ,F y ,F z Torque τ; rotary encoders, which measure rotation angle θ and angular velocity ω (angular velocity can also be calculated by differentiation); linear displacement sensors, which measure axial position z and velocity v (velocity can also be calculated by differentiation); tilt sensors, which measure tool posture.

[0030] Based on the raw measurement data, the sensing module can calculate the following state parameters in real time:

[0031] Screwing progress φ prog It represents the current ingress depth z and the target depth z. target The ratio. From 0 (just starting) to 1 (fully screwed in), it provides a time reference for the entire screwing-in process.

[0032] The screw-in energy ratio ξ represents the ratio of power consumed in the rotational direction to power consumed in the axial thrust. Ideally, this ratio should be close to a constant determined by the screw's geometry. In actual calculations, the system normalizes the original ratio to obtain a standardized state indicator. A positive and large indicator indicates that rotation is more difficult than expected, possibly due to resistance; a negative indicator indicates that rotation is too easy, potentially leading to stripping; and a value close to zero indicates normal screw-in. Specifically, the original value ξ of the screw-in energy ratio parameter is first calculated. raw The original ratio is then normalized. It's important to note that the normal baseline value for the infiltration energy ratio needs to be corrected using a real-time estimated friction coefficient to make the parameter more accurate and avoid misjudgments due to differences in material friction characteristics. The infiltration energy ratio ξ is specifically: ξ(t) = (ξ... raw (t)-ξ nom,adj (t)) / ξ nom,adj (t), ξ raw (t)=(τ(t)·ω(t)) / (F z (t)·v(t)+ε), In the formula, the appended (t) after the parameter indicates that the parameter is a time-varying parameter, ξ nom The nominal screw-in energy ratio is determined by the thread geometry; ε is a numerically stable term to prevent the denominator from being zero; κ fric The friction correction coefficient is typically taken as 0.2-0.5, μ. nomThe friction coefficient is designed, and μ is the composite friction coefficient estimated in real time by the subsequent observer.

[0033] Radial deviation e radial This indicates the degree of offset between the screw center and the threaded hole center. The lateral force F is obtained through measurement. x ,F y The contact stiffness k obtained from offline calibration or online identification contact The system calculates the actual offset distance and direction. Specifically, the radial deviation e radial for:

[0034] , where θ tool This is the current azimuth angle of the tool.

[0035] Contact Normal: By analyzing three-dimensional force measurements and thread profile angles, the system calculates the magnitude and direction of the normal force on the contact surface. This helps determine if the screw is tilted and the direction and angle of tilt (e.g., ...). Figure 2 As shown, the screw and the threaded hole are tilted (i.e., the screw axis and the central axis of the threaded hole are not completely coincident, but have a certain angle), which provides key information for automatic alignment. Specifically, the angle Φ between the screw axis and the contact normal is... contact for:

[0036] , where φ thread This refers to the thread profile angle.

[0037] The engagement confidence level, c, represents the probability of proper thread engagement. The system analyzes the correlation between changes in axial force and rotation angle: a strong correlation (i.e., axial force increases synchronously during tightening) indicates good engagement and high confidence; a weak correlation indicates potential stripping or freewheeling and low confidence. Specifically, the engagement confidence level c is: Where N is the sliding window length, F is the sampling frequency multiplied by the time it takes to rotate one revolution, i.e., the number of sampling points in one revolution, and F z ' is the axial force F within the window z The mean of , θ' is the mean of the rotation angle θ within the window, β is the gain coefficient, usually taken as 10-20, ρ thresh The threshold is typically set between 0.6 and 0.8.

[0038] In addition to directly calculating parameters, the system also uses a state observer to estimate a key internal state that cannot be directly measured—the friction coefficient μ. The system uses a model-based observer to estimate the static and dynamic friction coefficients of the threaded contact surface in real time. The observer uses measured torque, axial force, and screw geometry parameters, continuously updating the friction coefficient estimate by comparing the actual torque with the theoretically predicted torque. In this way, the system can sense changes in the thread friction state: a sudden increase in the friction coefficient may indicate the presence of foreign objects; a persistently low coefficient may indicate thread wear; and abnormal fluctuations may indicate thread damage. The observer is designed as follows:

[0039] Where, the symbol ^ represents the estimated value, μ s μ is the static friction coefficient. d μ is the coefficient of kinetic friction. s0 and μ d0 All are prior values ​​of the friction coefficient, α μ K is the forgetting factor, typically taken as 0.01-0.1. μ1 and K μ2 All are observer gains, K robust For robust gain, r eff Let λ be the effective radius of the screw, λ be the helix angle, and sat be the saturation function, sat(x) = min(1, max(-1, x)). ω Let be the angular velocity gain, which can be 10. Therefore, the overall friction coefficient μ can be obtained.

[0040] Contact State Classification: Based on the estimated friction coefficient μ and engagement confidence c, the system uses a fuzzy logic classifier to classify the current engagement state into four modes: free engagement (screw not yet in contact), initial engagement (starting to contact but not fully engaged), normal engagement (stable engagement state), and abnormal friction (abnormal resistance). The classifier outputs the most probable state mode based on the membership degree of the input parameters and preset rules, providing high-level guidance for subsequent decision-making. The input membership function is as follows:

[0041] , where μ thresh μ is the classification threshold for the friction coefficient. min μ is the minimum friction coefficient. max c is the maximum value of the friction coefficient. thresh c is the classification threshold for engagement confidence. min c is the minimum confidence level of meshing. max The above value represents the maximum confidence level of engagement and can be obtained through offline calibration.

[0042] The rule base is:

[0043] Rule 1: If μ is low (μ is less than μ) lowAnd c is low (c is less than c) low If ), then the mode = free spin-in (FREE);

[0044] Rule 2: If μ is in (μ in μ) low and μ high (between) and c is in (c in c) low and c high If the engagement is between (the two), then the mode = initial engagement (INITIAL).

[0045] Rule 3: If μ is in (μ in μ) low and μ high (between) and c is high (c is greater than c) high If ), then the mode = normal engagement.

[0046] Rule 4: If μ is high (μ is greater than μ) high And c is low (c is less than c) low If the pattern is abnormal friction (ABNORMAL), then the pattern is abnormal friction.

[0047] II. Dynamic Helical Field Generator: Based on the current state, dynamically adjust the helical field parameters to generate the desired axial trajectory, axial force, and rotational speed.

[0048] Traditional spiral control uses fixed trajectories and fixed force commands, which cannot adapt to dynamically changing spiral conditions. This invention creatively creates a dynamically adjustable virtual spiral field, which can adjust key parameters in real time according to the spiral state, generating a smooth, efficient, and adaptive spiral strategy, and providing intelligent expected commands for the control layer.

[0049] Based on the sensed state, the system dynamically sets three key parameters of the virtual spiral field:

[0050] Adaptive pitch h: The pitch determines the distance traveled per revolution. The system dynamically adjusts the pitch based on the screw-in progress and screw-in energy ratio. The basic strategy is: as the screw-in depth increases, the pitch is gradually reduced to achieve fine control; when the screw-in energy ratio indicates high resistance, the pitch is further reduced to enter fine control mode; when the resistance is low, the pitch is appropriately increased to improve efficiency. All adjustments are smooth transitions, avoiding abrupt changes. Specifically, the adaptive pitch h is:

[0051] Where h0 is the nominal pitch of the screw, α φ β is the progress contraction factor, typically taken as 0.1-0.3. φ The rate of rate of progress contraction is usually taken as 3-5, α ξ This is the state adjustment amplitude coefficient, typically taken as 0.05-0.15, β ξ This is the state adjustment rate coefficient, typically taken as 2-4.

[0052] Adaptive helix radius R: The helix radius determines the thickness of the screw-in path, used to achieve automatic centering. Its base radius naturally decreases as the screw-in progresses, conforming to physical laws. Simultaneously, the system calculates a correction term based on radial deviation: when screw misalignment is detected, the helix path actively shifts in the opposite direction, guiding the screw back to the center. Importantly, the correction intensity decreases as the screw-in progresses, with active correction in the initial stage and reduced intervention in the later stages to avoid wobbling in the final stage. The helix radius R is specifically:

[0053] Where R0 is the initial helix radius, typically taken as 1.5-2 times the screw radius, R min The minimum helix radius is typically taken as 1.1-1.2 times the screw radius, δ max For maximum correction gain, it is typically taken as 0.5-1.0 mm / N, δ min The minimum correction gain is typically set to 0.1-0.2 mm / N.

[0054] Based on the above virtual spiral field, the specific desired control value can be calculated:

[0055] Desired axial force F zd Desired axial force F zd The magnitude of the force is intelligently adjusted based on the insertion state. The basic trend is to increase with increasing depth to overcome the increased resistance. When the insertion energy ratio indicates abnormally high resistance, the desired axial force F is further increased. zd To help overcome resistance; when the resistance is abnormally low or the engagement confidence is low, reduce the desired axial force F. zd To prevent damage to the threads. Furthermore, different contact modes correspond to different force coefficients; for example, a smaller force is used during free screwing, while a larger force, but with greater caution, is used during abnormal friction. Specifically, the desired axial force F... zd for:

[0056]

[0057] Among them, F z0 The initial axial force, γ, is set according to the screw specifications. φ γ is the schedule adjustment factor, typically taken as 0.5-1.0. ξ This is the state adjustment coefficient, which is usually taken as 0.2-0.5.

[0058] Desired axial trajectory z d This value represents the target depth over time, calculated by subtracting the advance per revolution based on the adaptive pitch from the initial position. Additionally, if screw tilt is detected, the trajectory includes a periodic fine-tuning term to compensate for depth changes caused by the tilt. Specifically: Where z0 is the initial axial position, kz This is the tilt compensation coefficient.

[0059] Desired angular velocity ω d : Desired rotational speed ω d The rotation speed is determined by multiple factors, including the base speed, which gradually decreases as the rotation progresses. Simultaneously, a coupling mechanism is introduced: when the actual axial force falls short of the desired value, the rotation speed is appropriately reduced to allow the thrust to catch up; when the axial force exceeds the desired value, the rotation speed is appropriately increased. The coupling strength itself is also adaptive: when friction or resistance is high, the coupling strength is reduced to avoid overreacting to small force errors. Specifically, the desired angular velocity ω... d for:

[0060] Where ω0 is the nominal angular velocity, set according to process requirements, and η φ For the schedule adjustment coefficient, η0,μ η ,ν η The coupling gain parameter can be determined experimentally.

[0061] III. Compliant Coupling Controller: Based on the desired command and actual feedback, the controller calculates the axial acceleration command and rotational torque command, and sends the control commands to the motor driver and axial actuator. The controller parameters are adjusted in real time according to the status.

[0062] Based on the desired instructions generated by the dynamic spiral field generator and combined with real-time feedback, precise control commands are calculated to drive the robot's movements, ensuring smooth and safe motion while adjusting control characteristics according to the spiraling state. This module achieves high-precision, highly compliant trajectory tracking and force control, ensuring a stable and reliable spiraling process, and intelligently adjusting the control strategy according to the spiraling state, exhibiting excellent dynamic performance and robustness.

[0063] (a) Axial compliance control

[0064] Axial control employs admittance control. M z It is the axial virtual inertia, e z For axial position error, unlike traditional admittance control, the key feature of the axial compliance control in this invention is adaptive parameter adjustment, which specifically includes:

[0065] Time-varying stiffness K zThe system's "stiffness" is adjusted in real time according to the screwing state. As the screwing progresses, the stiffness gradually increases, making control more precise. When the screwing energy ratio shows an anomaly (regardless of whether the resistance is too high or too low), the stiffness is increased for more decisive action. When the engagement confidence is low, the stiffness is decreased for smoother action, avoiding damage to the threads. Different contact modes also correspond to different stiffness coefficients; for example, it is softer during free screwing and harder but more cautious during abnormal friction. The time-varying stiffness is specifically as follows: , where K z0 Based on the fundamental stiffness, and designed according to the system inertia and bandwidth requirements, λ φ λ is the schedule stiffness coefficient, typically taken as 0.5-1.0. ξ This is the state stiffness coefficient, which is usually taken as 0.2-0.5.

[0066] Time-varying damping B z To maintain a system response that is both fast and smooth, damping parameters are calculated in real time based on stiffness and system mass to ensure proximity to the critical damping state. Specifically: , where ζ is the damping ratio.

[0067] The axial acceleration command is: .

[0068] (ii) Rotational compliance control

[0069] The rotation control employs a feedforward-feedback composite structure. The feedback section uses a PID controller to track the desired angle; the feedforward section includes inertia compensation and friction compensation. Specifically:

[0070] Among them, e θ Let J be the rotational angular error, J be the moment of inertia, and τ be the rotational angular error. c0 τ is the nominal Coulomb coefficient of friction. c1 For static friction estimation gain, B0 is the nominal viscous friction coefficient, B1 is the kinetic friction estimation gain, and K... p0 ,K i0 ,K d0 Based on the PID gain, obtained through system identification, κ pφ ,κ iφ ,κ dφ All are progress adjustment coefficients, κ pξ ,κ iξ ,κ dξ All are state adjustment coefficients.

[0071] Specifically, the observation-based friction feedforward compensation term uses the estimated friction coefficient to pre-calculate and compensate for the friction torque. This significantly improves tracking accuracy and reduces steady-state errors caused by friction.

[0072] The parameters of a PID controller are not fixed but adjusted according to the input state. For example, the proportional gain increases with the input state to improve response speed; the integral gain decreases with the input state to avoid overshoot; and the derivative gain is adjusted according to the input state to enhance the damping effect. In this way, the controller exhibits different characteristics at different input stages to adapt to different needs.

[0073] Alternatively, impedance control can be used as the core of the rotary controller because axial compliance control already employs admittance control (force controls position), and impedance control (position controls force) can complement and coordinate well with it. It should be noted that the rotational direction here primarily involves position tracking (angle) while also requiring compliance (torque limitation). Therefore, an impedance-based rotary controller can adjust impedance parameters according to the rotational state and coordinate with axial admittance control. Axial control uses admittance control (position adjustment generated by force error), while rotation uses impedance control (torque command generated by position error). In this way, both can be unified into interactive force and position control, and the dynamics of the two directions can be dynamically linked through coupling terms, thereby achieving cross-coupled impedance control.

[0074] Specifically, the rotational impedance control law is as follows:

[0075] , where τ cmd This refers to the output torque control command, e θ For angular error, M θ It is the virtual inertia in the direction of rotation, B θ and K θ It is time-varying damping and stiffness, adjusted according to the screw-in state, τ ext It is the external torque (the deviation between the actual measured torque and the model-predicted torque), τ couple This is the torque term resulting from the coupling of axial force to rotation, ζ is the damping ratio, typically taken as 0.7-1.0, and K... θ0 The basic value for rotational stiffness is selected based on the desired rotational stiffness and can be determined experimentally to ensure that the angle error is within an acceptable range during normal rotation. θφ ,κ θξ All are influence coefficients, κ θφ Adjust the stiffness according to the effect of the screwing progress; generally, the stiffness increases with the increase of the screwing progress, and can be taken as 0.5-1.0; κ θξ Adjustments are made based on the energy ratio deviation; when the resistance is high, the stiffness is increased, and a value of 0.3-0.8 can be taken. η0 is the basic coupling gain, determined experimentally, to ensure appropriate torque compensation caused by axial force error during normal screwing. μ η ,ν η All are adjustment coefficients, determined experimentally, τ fric,ff It is based on feedforward compensation from friction observations, the same as before.

[0076] Alternatively, the impedance parameter can be adjusted using a relative ratio strategy, specifically:

[0077] , of which F z,nom The nominal axial force represents the axial force expected to be applied under ideal standard operating conditions. This adjustment method offers greater adaptability and safety, making it more suitable for scenarios with uncertainties and disturbances, such as thread hole machining errors, changes in material hardness, and changes in lubrication conditions, as well as scenarios with high safety requirements, such as preventing thread stripping and avoiding overtightening.

[0078] This invention also discloses a working method for a robot screw-tightening system based on dynamic helical field and compliance control, comprising:

[0079] Sensing: Reading sensor data, including axial force F z Calculate the screw-in progress φ using torque τ, rotation angle θ, angular velocity ω, axial position z, and velocity v. prog Injection energy ratio ξ, radial deviation e radial Contact normal angle Φ contact The system calculates the engagement confidence level c and estimates the friction coefficient μ of the thread contact surface in real time using a state observer. Based on the friction coefficient μ and engagement confidence level c, the system classifies the current contact state.

[0080] Dynamic spiral field generation: Based on the state parameters output by the sensing module, the adaptive pitch h and adaptive spiral radius R of the virtual spiral field are dynamically adjusted, and the desired axial force F is generated. zd Desired axial trajectory z d and desired angular velocity ω d ;

[0081] Compliant coupling control: receiving the desired axial force F zd Desired axial trajectory z d and desired angular velocity ω d And combined with the real-time feedback of axial force F z Given the position z, an axial acceleration command is calculated using a parameter-adaptive axial admittance control law; simultaneously, a rotational torque command is calculated using a parameter-adaptive rotational control law, which is either a feedforward-feedback composite control law or a rotational impedance control law.

[0082] The system of the present invention can be used to implement the aforementioned method.

[0083] Compared with traditional screw-in control methods, this invention can achieve: deep state perception, which not only uses direct measurement values ​​but also estimates key internal states through state observers, providing a more comprehensive state understanding; dynamic trajectory generation, which abandons fixed trajectories and dynamically adjusts the helical field parameters according to real-time state, achieving true adaptive control; intelligent coordinated control, where axial control and rotational control are not independent but work in coordination through an adaptive coupling mechanism, mimicking the natural action of humans tightening screws; and multi-mode adaptation, where the system automatically switches control strategies based on state classification in different screw-in stages and under abnormal conditions.

[0084] This invention can be widely applied to various scenarios requiring precision screw insertion. For example, in the aerospace field, where reliability and accuracy requirements are extremely high, the system can use conservative parameters. In the automotive manufacturing field, where high efficiency and mass production are pursued, the system can optimize the insertion speed and ensure quality consistency through intelligent adaptation. In the field of electronic product assembly, where screws are tiny and require high precision, the system can improve sensing accuracy and control frequency to achieve fine operation. In maintenance scenarios, when faced with potentially damaged threads and unknown conditions, the system can use exploratory parameters to identify thread conditions through trial actions and select an appropriate strategy.

Claims

1. A robotic screw-tightening system based on dynamic helical field and compliant control, characterized in that, include: The sensing module is used to read sensor data, including axial force F. z Calculate the screw-in progress φ using torque τ, rotation angle θ, angular velocity ω, axial position z, and velocity v. prog Injection energy ratio ξ, radial deviation e radial Contact normal angle Φ contact The system calculates the engagement confidence level c and estimates the friction coefficient μ of the thread contact surface in real time using a state observer. Based on the friction coefficient μ and engagement confidence level c, the system classifies the current contact state. The formula for calculating the inversion energy ratio is: ξ(t) = (ξ raw (t)-ξ nom,adj (t)) / ξ nom,adj (t), ξ raw (t)=(τ(t)·ω(t)) / (F z (t)·v(t)+ε), , where ξ nom ε is the nominal inclination energy ratio, κ is the numerical stability term, and κ is the nominal inclination energy ratio. fric μ is the friction correction coefficient. nom To design the coefficient of friction; The formula for calculating engagement confidence is: Where N is the length of the sliding window, F z ' is the axial force F within the window z The mean of , θ' is the mean of the rotation angle θ within the window, β is the gain coefficient, ρ thresh For the threshold; The dynamic spiral field generator is used to dynamically adjust the adaptive pitch h and adaptive spiral radius R of the virtual spiral field based on the state parameters output by the sensing module, and to generate the desired axial force F. zd Desired axial trajectory z d and desired angular velocity ω d ; Compliant coupling controller for receiving the desired axial force F zd Desired axial trajectory z d and desired angular velocity ω d And combined with the real-time feedback of axial force F z Given the position z, an axial acceleration command is calculated using a parameter-adaptive axial admittance control law; simultaneously, a rotational torque command is calculated using a parameter-adaptive rotational control law, which is either a feedforward-feedback composite control law or a rotational impedance control law.

2. The robot screw-tightening system according to claim 1, characterized in that, The state observer in the sensing module is used to estimate the static friction coefficient μ. s and the coefficient of kinetic friction μ d Its update law is based on the measured torque, axial force, screw geometry parameters, and the difference between the theoretically predicted torque and the actual torque. The state observer outputs the comprehensive friction coefficient μ.

3. The robot screw-tightening system according to claim 1, characterized in that, The perception module uses a fuzzy logic classifier, taking the estimated friction coefficient μ and engagement confidence c as input, to classify the contact state into four modes: free engagement, initial engagement, normal engagement, and abnormal friction.

4. The robot screw-tightening system according to claim 1, characterized in that, In the dynamic helical field generator, the formula for calculating the adaptive pitch is: Where h0 is the nominal pitch of the screw, α φ ,β φ ,α ξ ,β ξ This is for adjusting the coefficient.

5. The robot screw-tightening system according to claim 1, characterized in that, In the dynamic spiral field generator, the formula for calculating the adaptive spiral radius R includes a correction term based on radial deviation, and the correction gain decreases as the spiraling progress increases.

6. The robot screw-tightening system according to claim 1, characterized in that, The axial admittance control law in the compliant coupling controller is: , of which M z e represents the axial virtual inertia. z For axial position error, stiffness K z and damping B z According to the screwing progress φ prog The spin-in energy ratio ξ and engagement confidence c are adaptively adjusted.

7. The robot screw-tightening system according to claim 1, characterized in that, The rotational control law in the compliant coupling controller adopts the rotational impedance control law: , of which M θ It is the virtual inertia in the direction of rotation, B θ and K θ It is time-varying damping and stiffness, τ ext It is the external torque, stiffness K θ According to the screwing progress φ prog Adaptive adjustment of the engagement energy ratio ξ and meshing confidence c; coupling torque term τ couple Related to the axial force error, its coupling gain η is adaptively adjusted based on the spin-in energy ratio ξ and the friction coefficient μ.

8. A method for operating a robot screw-tightening system based on dynamic helical field and compliant control as described in any one of claims 1-7, characterized in that, include: Sensing: Reading sensor data, including axial force F z Calculate the screw-in progress φ using torque τ, rotation angle θ, angular velocity ω, axial position z, and velocity v. prog Injection energy ratio ξ, radial deviation e radial Contact normal angle Φ contact The system calculates the engagement confidence level c and estimates the friction coefficient μ of the thread contact surface in real time using a state observer. Based on the friction coefficient μ and engagement confidence level c, the system classifies the current contact state. The formula for calculating the inversion energy ratio is: ξ(t) = (ξ raw (t)-ξ nom,adj (t)) / ξ nom,adj (t), ξ raw (t)=(τ(t)·ω(t)) / (F z (t)·v(t)+ε), , where ξ nom ε is the nominal inclination energy ratio, κ is the numerical stability term, and κ is the nominal inclination energy ratio. fric μ is the friction correction coefficient. nom To design the coefficient of friction; The formula for calculating engagement confidence is: Where N is the length of the sliding window, F z ' is the axial force F within the window z The mean of , θ' is the mean of the rotation angle θ within the window, β is the gain coefficient, ρ thresh For the threshold; Dynamic spiral field generation: Based on the output state parameters, the adaptive pitch h and adaptive spiral radius R of the virtual spiral field are dynamically adjusted, and the desired axial force F is generated. zd Desired axial trajectory z d and desired angular velocity ω d ; Compliant coupling control: receiving the desired axial force F zd Desired axial trajectory z d and desired angular velocity ω d And combined with the real-time feedback of axial force F z Given the position z, an axial acceleration command is calculated using a parameter-adaptive axial admittance control law; simultaneously, a rotational torque command is calculated using a parameter-adaptive rotational control law, which is either a feedforward-feedback composite control law or a rotational impedance control law.

Images