Intelligent power position cooperative control and precision guarantee method and system

Abstract

The invention relates to the technical field of industrial robot precision assembly, in particular to an intelligent power position cooperative control and precision guarantee method and system which are suitable for micron-level precision scenes such as semiconductor packaging, medical instrument precision assembly and aerospace component butt joint. Real-time sensing of the assembly force-position state, self-adaptive adjustment of parameters, dynamic optimization of a path and online verification of quality can be realized, and the problems of force-position coupling interference, difficulty in guaranteeing precision and easiness in damaging parts in traditional assembly are solved.

Description

Technical Field

[0001] This invention relates to the field of precision assembly technology for industrial robots, and in particular to an intelligent force-position coordination control and precision assurance method and system. Background Technology

[0002] Automated precision assembly using industrial robots is a core process in high-end manufacturing, and the accuracy of its force-position coordination control directly determines the performance and reliability of the product. Currently, the assembly process faces four major technical bottlenecks: First, force-position coupling interference: in traditional control, position and force control are independent, making it difficult to handle the dynamic coupling relationship between force and position during assembly, leading to accuracy fluctuations; Second, poor parameter adaptability: fixed force-position parameters cannot match the fitting characteristics (interference / clearance fit) of different parts, easily causing scratches or improper assembly; Third, weak path anti-interference capability: preset trajectories cannot cope with part positioning errors and flexible deformation, leading to assembly stalls; Fourth, delayed quality verification: relying on post-assembly inspection, it is impossible to correct force-position parameters in real time, causing batch quality risks.

[0003] In existing technologies, some solutions improve positioning accuracy through hardware upgrades, but lack intelligent decision-making based on force-position coordination; some solutions use simple impedance control, which can only adapt to rigid assembly scenarios; and some solutions rely on manual teaching, making it difficult to meet the needs of flexible production with multiple varieties and small batches. Therefore, there is an urgent need to build an intelligent assembly control system that integrates force-position coupling modeling and dynamic decision-making. Summary of the Invention

[0004] This invention provides an intelligent force-position collaborative control and accuracy assurance method and system. Through a four-layer architecture—six-dimensional force-vision-displacement fusion perception, force-position coupling dynamic modeling, intelligent collaborative control, and online path optimization—it achieves micron-level precision in force-position collaborative assembly. The core innovation lies in proposing four key algorithms: a force-position coupling-based dynamic assembly state recognition algorithm, an adaptive impedance-based force-position collaborative control algorithm, an error-compensated assembly path optimization algorithm, and a multi-dimensional assembly quality real-time verification algorithm. A complete mathematical model and efficiency enhancement mechanism are also established, significantly improving assembly accuracy and reliability.

[0005] A first aspect of this invention provides a method for intelligent force-position coordination control and accuracy assurance in the automated assembly process of an industrial robot, comprising the following steps:

[0006] Multi-source information acquisition during assembly process: Deploy a six-dimensional force-vision-displacement fusion sensing system to collect assembly contact force / torque, part three-dimensional pose, robot end position and part deformation data in real time;

[0007] Dynamic identification of assembly status: Based on the collected data, the dynamic identification of assembly status based on force-position coupling is realized to achieve accurate identification of the mating stage and the force-position coupling state;

[0008] Force-position coordinated control: Based on the state recognition results, the adaptive impedance force-position coordinated control algorithm is activated to dynamically optimize the force-position control parameters;

[0009] Path optimization and quality verification: An assembly path optimization algorithm with error compensation is used to generate an anti-disturbance trajectory, which is then combined with a multi-dimensional assembly quality real-time verification algorithm to judge the assembly effect.

[0010] Full-process closed-loop optimization: Based on historical data, establish a correlation model of accuracy, force control stability and damage rate, continuously iterate and optimize algorithm parameters, and achieve continuous improvement of the assembly process.

[0011] A second aspect of the present invention provides a system for implementing the above method, comprising:

[0012] Multi-source sensing units: six-dimensional force sensor (range ±200N, resolution ≤0.01N), high-precision vision system (resolution ≤1μm, frame rate ≥100fps), laser displacement sensor (accuracy ≤0.1μm);

[0013] Data processing unit: Industrial computer (CPU≥i9, GPU≥RTX4090), FPGA real-time processing module (control cycle≤1ms);

[0014] Intelligent decision-making unit: Deploys four core algorithms, integrated into the RT-Linux real-time operating system, to achieve force-position coordinated control and path optimization;

[0015] Robot control unit: six-axis precision assembly robot (repeat positioning accuracy ≤2μm), servo drive system (response bandwidth ≥1kHz);

[0016] Human-computer interaction unit: monitoring terminal (supports 3D force-position curve visualization), parameter configuration interface, and quality analysis report module. Attached image description:

[0017] Figure 1 The flowchart of the implementation principle of this system method.

[0018] Beneficial effects

[0019] Compared with the prior art, the present invention has the following beneficial effects:

[0020] Assembly accuracy is significantly improved: the force-position coordination control accuracy reaches ±3μm, and the fit clearance is controlled within ±2μm, meeting the micron-level assembly requirements of high-end manufacturing such as semiconductors and aerospace;

[0021] Assembly reliability has been greatly improved: the damage rate of parts has been reduced from 12% to less than 1%, the assembly success rate has been increased from 80% to more than 99%, and the consistency of mass production quality has been improved by 70%.

[0022] Production efficiency optimization: Number of stalls and retry attempts reduced by 88%, assembly cycle shortened by 35-45%, and equipment utilization rate increased by 40%;

[0023] Enhanced flexibility: The adaptive impedance parameters can be adapted to different mating scenarios with interference fits of 0-5μm, reducing changeover adjustment time by more than 65% and meeting the needs of flexible production of multiple varieties and small batches. Detailed Implementation

[0024] Example 1: Dynamic Recognition Algorithm for Assembly State Based on Force-Position Coupling

[0025] Algorithm principle: Construct a coupled correlation model of assembly force features and position features, and realize real-time identification of assembly stages (alignment / contact / insertion / completion) and abnormal states (jamming / off-center load) by extracting force-position coupling features (such as force change rate-displacement gradient correlation).

[0026] Key innovations:

[0027] Propose force-potential coupling eigenvectors:

[0028] C = [Fx,Fy,Fz,τx,τy,τz,Δx,Δy,Δz,F˙n / x˙,cov(Ft,Δθ)], where F˙n / x˙ is the ratio of normal force to displacement rate of change, and cov(Ft,Δθ) is the tangential force to attitude deviation covariance;

[0029] Establish a state recognition model: S = GNN-LSTM(C), capture the force-potential coupling relationship through graph neural network, and combine it with LSTM to realize temporal state classification;

[0030] Design a coupling state early warning index: Ic = ∑i = 111 αi·ci0 ci -ci0, triggering an anomaly warning when Ic > ∈.

[0031] Modeling efficiency enhancement principle:

[0032] Force-position coupling characteristics overcome the limitations of single force / position signals, establish a positive correlation model between recognition accuracy and assembly success rate, and improve the state recognition accuracy to over 98%.

[0033] Coupling status early warning detects anomalies 5-10ms in advance, avoiding component damage and reducing assembly failure rate by 80%.

[0034] The overall assembly status perception is comprehensive and forward-looking, and the robustness of status judgment in complex cooperation scenarios is significantly enhanced.

[0035] Innovation Point 2: Adaptive Impedance Force-Position Coordinated Control Algorithm

[0036] Algorithm principle: Establish an assembly force-position coupling dynamic model, and dynamically adjust the impedance parameters (stiffness K, damping B) based on real-time force-position feedback to achieve coordinated optimization of force and position, avoiding force overshoot and position deviation.

[0037] Key innovations:

[0038] A force-potential coupling dynamic model is proposed: Mx¨+Bx˙+Kx=Fd-Fe, where M is the equivalent mass, Fd is the desired force, and Fe is the actual contact force, thus quantifying the dynamic coupling relationship between force and potential.

[0039] Design an adaptive impedance parameter adjustment law: K=K0·f(F˙n,Δx), B=B0·g(cov(F,x)), and dynamically optimize the impedance parameters through the force change rate and position deviation;

[0040] Construct a force-position cooperative error model: e=λe·(xd-x)+(1-λe)·(Fd-Fe), where λe∈[0,1] is the cooperative weight, to achieve weighted optimization of force-position error.

[0041] Modeling efficiency enhancement principle:

[0042] The coupled dynamics model accurately describes the force-position interaction law, with force control error controlled within ±0.5N and position accuracy improved to ±3μm;

[0043] The adaptive impedance parameter improves the force-position adaptability of different mating scenarios (interference 0-5μm) by 70% and reduces the component damage rate by 90%.

[0044] The overall force-position coordination control accuracy is significantly improved, the force fluctuation amplitude during assembly is reduced by 60%, and the position tracking error is reduced by 50%.

[0045] Innovation Point 3: Assembly Path Optimization Algorithm with Error Compensation

[0046] Algorithm principle: A comprehensive error model is constructed based on part positioning error and flexible deformation data. An improved Bézier curve is used to optimize the path, thereby achieving real-time error compensation and improved path smoothness.

[0047] Key innovations:

[0048] Establish a comprehensive error model: ε=εpos+εdef+εsen, where εpos ​​is the positioning error, εdef is the deformation error, and εsen is the sensing error;

[0049] A path compensation model is proposed: P′(t)=P(t)-ε^(t)-kp·∫e(t)dt, which compensates for the cumulative error through integration;

[0050] Design the objective function for path optimization: J=∫||P¨(t)||2dt+μ·∫||ε(t)||2dt, to balance path smoothness and error compensation effect.

[0051] Modeling efficiency enhancement principle:

[0052] The comprehensive error model quantifies the impact of multi-source errors on assembly accuracy, and the assembly gap is controlled within ±2μm after error compensation.

[0053] Smooth path optimization reduces the impact of robot movement, further reduces force fluctuations by 30%, and extends equipment life by 20%.

[0054] The overall assembly path's resilience and accuracy have been improved, and the number of retries due to jamming has been reduced from an average of 2.5 times to 0.3 times.

[0055] Innovation Point 4: Multi-dimensional Assembly Quality Real-time Verification Algorithm

[0056] Algorithm principle: Construct a multi-dimensional verification system based on assembly force curve similarity, fit clearance, and fastening force. Through feature fusion and statistical analysis, realize real-time assessment of assembly quality and defect tracing.

[0057] Key innovations:

[0058] Propose a quality verification index:

[0059] Q = wf·Sim(Fact,Fstd) + wg·(1-g / glim) + wt·(1-∣t-tstd∣ / tstd), where Sim(·) is the force curve similarity, g is the fit clearance, and t is the fastening force;

[0060] Establish a quality confidence model: Pq=Φ(Q;μq,σq2), and quantify the probability of quality compliance using a normal distribution function;

[0061] Design a defect tracing mechanism: Use a random forest model T(Q,K,B) to associate quality features with control parameters to quickly locate the cause of defects.

[0062] Modeling efficiency enhancement principle:

[0063] Multi-dimensional verification overcomes the limitations of a single indicator, improving the accuracy of quality assessment to 99% and reducing the rework rate by 85%.

[0064] Real-time verification and traceability reduce parameter adjustment response time to 10ms and improve process optimization efficiency by 60%.

[0065] The overall assembly quality can be controlled and traced in real time, and the quality consistency of mass production is improved by 70%.

[0066] Example 2

[0067] Example 1: Semiconductor Packaging Chip Mounting Scenario – Force-Position Co-control Application Based on “Micro-Force Sensing-Pose Compensation”

[0068] Technical Solution Design Background

[0069] Semiconductor chip mounting is a typical scenario for micron-level precision assembly. The core requirement is to achieve "high-precision alignment + damage-free mounting." The technical challenges are concentrated in three aspects: the miniaturization of chip size leads to extremely low tolerance for pose deviations, with even slight offsets affecting packaging performance; the fragile chip materials (silicon-based, ceramic substrates, etc.) are extremely sensitive to assembly forces, and excessive force can easily cause cracking and pin deformation; during mounting, force and pose are strongly coupled and interfere with each other, with even slight pose deviations causing a sudden increase in mounting force, and improper force control exacerbating pose deviations, creating a vicious cycle. Traditional assembly relies on fixed parameters and mechanical trajectories, lacking dynamic adaptation to the force-position coupling relationship, making it difficult to balance accuracy and safety. This embodiment innovatively constructs a "micro-force-pose dual closed-loop collaborative algorithm," a "multi-source interference adaptive suppression algorithm," a "mounting path dynamic correction algorithm," and a "micro-damage online monitoring algorithm." Through the collaborative efforts of these four algorithms, the force-position coupling problem is solved, achieving high precision, low damage, and high stability in chip mounting.

[0070] Core innovative algorithm principles and efficiency enhancement principles

[0071] Microforce-pose dual closed-loop collaborative algorithm

[0072] Algorithm Principle: Breaking through the limitations of traditional single-dimensional control, this algorithm constructs a dual-closed-loop architecture of "force feedback control loop + pose feedback control loop" to achieve dynamic balance between force and pose. The force control loop uses a high-precision force sensor to capture micro-force changes throughout the entire mounting process in real time, dividing the force threshold range into three levels: contact, bonding, and danger. Different control logics correspond to different ranges. The pose control loop dynamically captures the alignment deviation between the chip and the substrate using a vision sensor, generating micron-level compensation commands. The dual closed loops achieve synergy through synchronous data interaction: in the contact stage, pose calibration is the primary focus, with the force signal serving as the trigger threshold; in the bonding stage, force stability control is the primary focus, with the pose signal used for fine-tuning compensation; in the danger stage, force-position linkage braking is immediately triggered to avoid damage.

[0073] Enhancement Principle: The dual-closed-loop architecture completely eliminates force-position coupling interference, enabling force control and pose control to support each other rather than restrict each other. Compared to the traditional single-closed-loop architecture, the mounting alignment accuracy is significantly improved. At the same time, through dynamic adaptation of force thresholds, chip damage caused by rigid contact is avoided, achieving a dual guarantee of "accuracy and safety".

[0074] Multi-source interference adaptive suppression algorithm

[0075] Algorithm Principle: To address interference factors in industrial environments such as environmental vibration, equipment thermal drift, and sensor noise, a comprehensive processing mechanism of "interference feature extraction - classification modeling - dynamic suppression" is established. Multi-dimensional interference data is collected through accelerometers, temperature sensors, and signal filtering modules, extracting features such as vibration frequency, temperature drift rate, and signal noise amplitude. Based on unsupervised learning algorithms, interference is categorized into three types: high-frequency vibration, low-frequency drift, and random bursts. Influence models on force-position control are constructed for different interference types. Suppression strategies are matched for different interference types: high-frequency vibration is compensated for using a combination of damping adjustment and signal filtering; low-frequency drift is compensated for in real-time through temperature-pose mapping; and random bursts of interference trigger a temporary pause and recalibration process.

[0076] Enhancement Principle: The algorithm actively identifies and dynamically cancels complex on-site interference, avoiding misjudgments in force and position control caused by interference signals, and significantly improving system robustness. Compared with traditional passive anti-interference design, the force and position fluctuation amplitude during the placement process is greatly reduced, and the accuracy and stability remain consistent even in complex environments.

[0077] Dynamic correction algorithm for mounting path

[0078] Algorithm Principle: Abandoning the fixed trajectory planning model, the mounting path is dynamically optimized based on real-time force-position feedback. The mounting process is divided into three stages: alignment, contact, and bonding, each with preset initial parameters (speed, step distance, force threshold). During actual mounting, the path parameters are corrected in real time by using the force distribution uniformity captured by the force sensor and the alignment deviation captured by the vision sensor: in the contact stage, the approach speed is adjusted according to sudden changes in force signals to avoid rigid impact; in the bonding stage, the path direction is finely adjusted according to differences in force distribution to avoid local stress concentration; through pre-simulation analysis of chip damage-prone areas, path force-avoidance inflection points are set to avoid direct force on high-risk areas.

[0079] Efficiency enhancement principle: Dynamic path transforms the placement process from "mechanical execution" to "adaptive adaptation". The path parameters are matched with the chip status and force feedback in real time, which not only ensures alignment accuracy but also reduces local stress concentration, lowers the risk of chip breakage and pin deformation, and shortens the time of ineffective adjustment, thus improving placement efficiency.

[0080] Micro-damage online monitoring algorithm

[0081] Algorithm Principle: A multi-dimensional damage feature fusion and recognition model is constructed to achieve real-time perception of micro-damage during the mounting process. Visual sensors capture surface features such as cracks, chipping, and pin bending on the chip; force sensors analyze mechanical characteristics such as sudden changes in mounting force and fluctuation frequencies; and acoustic sensors collect abnormal acoustic signals during the mounting process (such as specific frequency sound waves generated by chip breakage). A feature-level fusion strategy is employed to map the data from the three types of sensors to a unified feature space. A classification model distinguishes between three states: normal mounting, minor damage, and severe damage, triggering corresponding processing mechanisms (continue normal mounting, mark and trace minor damage, and shut down urgently for severe damage).

[0082] The efficiency enhancement principle: The algorithm shifts from "post-event detection" to "in-process prevention," preventing damaged chips from flowing into subsequent processes and reducing rework costs and quality risks. Multi-dimensional feature fusion avoids the limitations of single sensor misjudgments, significantly improving the accuracy of micro-damage identification, while real-time early warning provides a basis for parameter optimization.

[0083] Implementation steps

[0084] Before placement, the industrial robot is equipped with a high-precision vision module, a six-dimensional force sensor, an acoustic sensor, and a temperature sensor. Coordinate system calibration with the semiconductor packaging worktable is performed to ensure spatial consistency of sensor data. Basic information about the chip and substrate is acquired through visual scanning, including chip size, pin distribution, and substrate pad positions. The three-segment parameters of the placement path are initialized (alignment speed, contact force threshold, and bonding force stability range). Initial environmental data is collected to establish an interference baseline, providing a reference for subsequent interference suppression.

[0085] The force-position dual-loop collaborative mounting robot drives the chip to approach the substrate according to the initial path. After entering the alignment section, the vision module captures alignment deviations in real time and dynamically adjusts the robot's posture to ensure initial alignment between the chip and the substrate pads. When the force sensor detects that a micro-force reaches the contact threshold, the dual-loop collaborative mode is triggered: the vision module switches to a high-precision acquisition mode to capture micron-level deviations and generate compensation commands; the force sensor continuously monitors force changes, and when the force value enters the bonding threshold range, the robot switches to a force stabilization control mode. Through small-step advancement and force feedback verification, it maintains the mounting force within a safe range. At the same time, the posture module fine-tunes the posture according to the force distribution data to ensure uniform bonding.

[0086] During the multi-source interference suppression and path correction mounting process, the multi-source interference adaptive suppression algorithm monitors environmental data in real time. When high-frequency vibration is detected, it automatically adjusts the robot joint damping coefficient and activates signal filtering to reduce the interference of vibration on the force and position signals. When thermal drift of the device is detected, the temperature-pose mapping model compensates for deviations caused by thermal deformation in real time. The path dynamic correction algorithm fine-tunes the mounting path direction based on force distribution uniformity data, avoiding local high-stress areas, ensuring uniform force on the chip, and preventing edge chipping and pin damage.

[0087] Micro-damage monitoring and process closure: The online micro-damage monitoring algorithm integrates visual, force, and acoustic data throughout the process to determine the placement status in real time. After placement, a high-precision visual scan of the chip surface is performed, and a placement quality report is generated by combining force signal fluctuation curves and acoustic characteristics. If the placement is determined to be normal, the force-position coordination parameters and alignment accuracy data are recorded to form a quality file; if minor damage is found, the chip number is marked and the placement parameters are traced; if severe damage is detected, an alarm is immediately triggered and the process is paused to avoid batch quality problems. The robot automatically picks up the next chip, adaptively adjusting the initialization parameters based on the previous chip's placement data to achieve closed-loop optimization of the process.

[0088] Qualitative explanation of efficiency enhancement

[0089] This embodiment significantly improves the overall performance of semiconductor chip mounting through four-layer algorithm collaboration: the force-position dual closed-loop collaboration greatly improves mounting alignment accuracy, completely solving the deviation problem caused by force-position coupling in traditional assembly; the multi-source interference suppression algorithm enhances the system's adaptability to complex on-site environments, ensuring that accuracy stability is not affected by environmental fluctuations; the path dynamic correction and micro-damage monitoring algorithms effectively reduce chip damage rate, reducing rework costs and quality risks; full-process automation and parameter adaptive optimization greatly reduce manual intervention, improve production efficiency and batch consistency, and fully meet the micron-level accuracy and high reliability requirements of semiconductor packaging.

[0090] Example 2: Precision Component Interlocking Scenario for Medical Devices – Collaborative Control Application Based on “Force-Position Decoupling-Path Pre-aiming”

[0091] Technical Solution Design Background

[0092] Precision component docking in medical devices (such as the connection between bushings and drive shafts of minimally invasive surgical instruments, and the sealing docking of implantable devices) is a typical "high-precision + high-reliability" assembly scenario. The technical challenges lie in: extremely small component clearances (micrometer level), stringent requirements for coaxiality and parallelism, where excessive deviation can affect sealing performance and movement flexibility; the component surfaces are often covered with biocompatible coatings or precision structures, requiring strict avoidance of scratches and deformation during assembly, necessitating extremely high control over assembly forces; and the force-position coupling exhibits nonlinear characteristics during docking, where even slight eccentricity can lead to an exponential increase in assembly force, easily causing component deformation or coating detachment. Traditional assembly relies on manual alignment and experience-based force control, which is insufficient to address the demands of nonlinear coupling and precision protection. This embodiment innovatively constructs a "force-position decoupling collaborative control algorithm," a "path pre-aiming-deviation compensation algorithm," a "coating protection force control algorithm," and an "online verification algorithm for sealing performance," achieving high precision, low damage, and high reliability in component docking through algorithmic collaboration.

[0093] Core innovative algorithm principles and efficiency enhancement principles

[0094] Force-position decoupling cooperative control algorithm

[0095] Algorithm Principle: To address the strong force-position coupling problem, a three-tiered processing mechanism of "modeling-decoupling-coordination" is proposed. First, through offline experiments and simulations, a nonlinear coupling model of "pose deviation-assembly force" is established to clarify the force variation patterns corresponding to different eccentricities and tilt angles. Second, an adaptive decoupling strategy is adopted, decomposing the control objective into independent pose control channels and force control channels, eliminating mutual interference between the two channels through a decoupling matrix. Finally, a collaborative feedback mechanism is established. After pose adjustment, the force change trend is predicted based on the coupling model, and the force control parameters are optimized in advance. When the force changes, the pose control strategy is corrected in reverse, achieving "decoupling without discreteness" collaborative control. When the force rise rate exceeds the safety threshold, the system automatically switches to "pose priority" mode, calibrates, and then resumes force-position coordination.

[0096] Enhancement Principle: The algorithm fundamentally solves the force-position coupling problem, eliminating the mutual constraints between pose accuracy and force stability. Compared to traditional coupling control, the coaxiality accuracy of component docking is significantly improved, while assembly force fluctuations are greatly reduced, avoiding component damage caused by coupling amplification and providing core technical support for precision docking.

[0097] Path Preview-Deviation Compensation Algorithm

[0098] Algorithm Principle: Based on the component's 3D model and mating requirements, a three-stage pre-aiming path of "introduction-alignment-advancement" is planned. The introduction stage uses conical guidance to reduce the difficulty of initial alignment, the alignment stage achieves micron-level calibration through high-precision sensing, and the advancement stage sets stable motion parameters. During the actual docking process, deviation data, including radial deviation, axial clearance, and angular tilt, are collected through the fusion of binocular vision and laser ranging sensors. A predictive control algorithm is used to anticipate the trend of deviation changes and generate compensation commands in advance. The compensation strategy adopts a "small step size, high frequency" mode, dynamically adjusting the compensation amplitude according to the magnitude of the deviation to avoid overshoot and oscillation, ensuring that the path always conforms to the optimal docking trajectory.

[0099] Efficiency Enhancement Principle: The anti-targeting mechanism transforms control from "passive response" to "active prediction," significantly improving the timeliness and accuracy of deviation compensation. Compared to traditional real-time feedback control, the jamming phenomenon during docking is completely eliminated, allowing components to advance smoothly even with minimal mating gaps. Simultaneously, it reduces ineffective adjustment actions, improving assembly efficiency and consistency.

[0100] Coating protection force control algorithm

[0101] Algorithm Principle: A "tiered force control + flexible buffering" protection mechanism is constructed to address the precision coatings and fragile structures on component surfaces. Force control modes are divided into three levels: exploration, protection, and emergency. The exploration mode is used for contact detection, employing low-stiffness force control to ensure gentle initial contact. The protection mode is the core operating mode, dynamically adjusting force control stiffness and movement speed to maintain assembly force within a safe range, avoiding rigid impacts. The emergency mode serves as a safety fallback, immediately triggering a shutdown and slight retraction when the force exceeds a threshold, preventing continuous high pressure from causing coating peeling or structural deformation. Simultaneously, a "force-speed" linkage logic is introduced, automatically reducing movement speed when the force approaches the threshold, minimizing impact effects through flexible propulsion.

[0102] Enhanced Performance Principle: A tiered force control and flexible buffering mechanism provide comprehensive protection for precision components, completely resolving coating scratches and structural deformation issues caused by traditional rigid force control. While ensuring docking accuracy, the surface integrity rate of components is significantly improved, meeting the stringent requirements of medical devices for biocompatibility and structural integrity.

[0103] Online verification algorithm for sealing performance

[0104] Algorithm Principle: A dual verification model based on "physical characteristics + functional indicators" is constructed to achieve real-time closed-loop verification of docking quality. Physical characteristic verification uses visual sensors to detect the mating gaps of the mating surfaces, ensuring there are no obvious gaps. Functional indicator verification uses pressure sensors to monitor the pressure stability within the sealed cavity, determining if there are any leaks. The dual verification data are fused to generate a quality judgment result. If the result is satisfactory, the process proceeds to the next step; otherwise, abnormal characteristics are automatically recorded and assembly parameters are traced, providing a basis for parameter optimization. The verification process is seamlessly integrated with the assembly process, requiring no additional inspection steps, achieving integrated "assembly-verification".

[0105] Efficiency Enhancement Principle: Online verification algorithms transform quality control from "post-production sampling" to "in-process full inspection," preventing defective products from entering subsequent processes, which is particularly suitable for high-reliability scenarios such as medical devices. The dual verification mechanism ensures the accuracy of quality judgment, while data traceability enables continuous optimization of assembly parameters, improving batch-to-batch consistency.

[0106] Implementation steps

[0107] Before docking, the industrial robot is equipped with a binocular vision module, laser rangefinder, six-dimensional force sensor, and air pressure sensor. Coordinate system calibration with the medical device assembly table is performed to ensure the consistency and accuracy of data from all sensors. Three-dimensional features of components (inner diameter of bushing, outer diameter of drive shaft, angle of mating conical surface, etc.) are acquired through visual scanning. Sealing performance requirements (air pressure holding threshold, fitting gap threshold) are input, and three-segment path parameters and force control threshold ranges are initialized. Based on the offline established force-position coupling model, decoupling matrices and collaborative control parameters are configured to provide a foundation for real-time control.

[0108] The force-position decoupling collaborative docking robot follows a pre-aiming path, driving the drive shaft into the bushing inlet section where a conical guide achieves initial alignment. Upon entering the alignment section, the force-position decoupling collaborative algorithm is activated. Vision and laser sensors collect pose deviation data, generating pose calibration commands through a decoupling matrix. Force sensors monitor assembly forces in real time; when the force reaches a contact threshold, a protection mode is activated, maintaining force stability through force-controlled stiffness adjustment. During docking, the algorithm dynamically monitors the force-position coupling state. When an abnormal force rise rate is detected, advancement is automatically paused, prioritizing pose calibration. Force-position collaborative advancement resumes only after the deviation is corrected, ensuring no amplified coupling interference throughout the process.

[0109] Path correction and coating protection Path preview - deviation compensation algorithm captures the deviation changes in real time during the docking process, predicts the development trend of the deviation, and generates micron-level compensation instructions. By making small-step adjustments, the component is always advanced along the optimal path to avoid jamming and eccentric wear. The coating protection force control algorithm maintains the protection mode throughout the process, dynamically adjusts the movement speed according to the force value changes, automatically decelerates when the force value approaches the threshold, and reduces the impact through flexible propulsion. The vision sensor monitors the surface state of the component in real time to ensure no problems such as coating scratches and structural deformations.

[0110] Sealing performance verification and process completion After docking, the online sealing performance verification algorithm is started: inject a set air pressure into the sealing cavity, the air pressure sensor continuously monitors the stability, and at the same time the vision sensor detects the fitting gap of the docking surface. After the dual verification data is fused, if it is judged to be qualified, record the force-position coordination parameters, deviation compensation data, and sealing verification results to form a complete quality file; if it is judged to be unqualified, automatically record the abnormal characteristics (such as air pressure leakage, excessive gap), trace the corresponding assembly parameters and trigger an alarm, and drive the robot to transfer the component to the repair area. The robot automatically grabs the next set of components, optimizes the initialization parameters based on the docking data of the previous set, and realizes the closed-loop iteration of the assembly process.

[0111] Efficiency qualitative description In this embodiment, through the collaboration of four-layer algorithms, the core technical problems of the docking of precision components in medical devices are effectively solved: the force-position decoupling and coordination algorithm completely eliminates the coupling interference, greatly improving the key precision indicators such as coaxiality; the path preview - deviation compensation algorithm ensures smooth docking under extremely small fitting gaps, avoiding jamming and eccentric wear; the coating protection force control algorithm realizes the full protection of the precision surface, meeting the structural integrity requirements of medical devices; the online sealing performance verification algorithm realizes real-time quality control and improves product reliability. Compared with the traditional assembly method, this solution not only greatly improves the assembly accuracy and quality stability, but also significantly reduces the manual intervention and rework costs, fully meeting the strict requirements of precision assembly in medical devices.

Claims

1. A method for intelligent force-position collaborative control and precision assurance in the automated assembly process of an industrial robot, characterized in that, include: Collect assembly contact force / torque, part 3D pose, robot end effector position, and part deformation data; The assembly state dynamic identification algorithm based on force-position coupling is used to achieve accurate identification of the mating stage and the force-position coupling state. The algorithm includes a force-position coupling feature vector C and a time sequence state identification model. The force-position coordination control algorithm based on adaptive impedance dynamically optimizes stiffness K and damping B to achieve force-position coordination control. The algorithm includes a coupled dynamics model and an adaptive parameter adjustment law. An assembly path optimization algorithm with error compensation is used to generate an anti-disturbance trajectory, and a multi-dimensional assembly quality real-time verification algorithm is used to judge the assembly effect. A correlation model of accuracy, force control stability, and damage rate is established based on historical data, and the algorithm parameters are continuously iterated and optimized to form a closed-loop control.

2. The method according to claim 1, characterized in that, The assembly state dynamic identification algorithm based on force-position coupling is designed with a coupling state early warning index I. c This enables early warning of abnormal conditions.

3. The method according to claim 1, characterized in that, The adaptive impedance force-potential collaborative control algorithm constructs a force-potential collaborative error model e = λ. e ·(x d -x)+(1-λ e )·(F d -F e This enables weighted optimization of force position error.

4. The method according to claim 1, characterized in that, The assembly path optimization algorithm for error compensation establishes a comprehensive error model ε = ε pos +ε def +ε sen This enables multi-source error quantization.

5. The method according to claim 1, characterized in that, The multi-dimensional assembly quality real-time verification algorithm proposes a quality verification index Q to achieve comprehensive evaluation of assembly quality.

Images