Self-adaptive clamp adjusting method based on intelligent vision and pneumatic control

Abstract

The invention discloses a self-adaptive clamp adjusting method based on intelligent vision and pneumatic control. The self-adaptive clamp adjusting method is suitable for a multi-variety small-batch flexible manufacturing scene. According to the method, a closed-loop control framework composed of multi-mode sensing, hybrid modeling, dynamic optimization and pneumatic execution is constructed, and precise sensing and rapid compensation of part deformation under the complex working condition are achieved. Submillimeter deformation data is obtained in real time by fusing binocular vision, optical fiber strain and a laser displacement sensor; dynamically outputting a clamp compensation instruction in combination with the finite element stiffness matrix and an XGBoost prediction model; a PSO-Adapt ivePena lty algorithm is introduced to carry out multi-objective optimization, so that the response speed and the cooperation efficiency of the system are improved; and the pneumatic actuating mechanism adopts a double-acting cylinder and a cam-groove mechanism, so that high-precision adjustment is realized. The method has the advantages of high precision, fast response and strong multi-clamp collaboration, and is suitable for the intelligent manufacturing fields of automobiles, aerospace and the like.

Description

Technical Field

[0001] The present invention relates to the field of intelligent manufacturing technology, and in particular to an adaptive fixture adjustment method based on intelligent vision and pneumatic control, which belongs to the cross-application technology of industrial automation, flexible assembly and multimodal perception control. Background Art

[0002] With the development of intelligent manufacturing, flexible production modes with multiple varieties and small batches have gradually become popular in industrial assembly, which puts higher requirements on the accuracy, response speed and adaptability of the fixture system.

[0003] Traditional fixtures are mostly fixed structures, making them difficult to adapt to complex workpieces or dynamic deformation scenarios. This is especially true when clamping thin-walled parts, which can easily cause excessive deformation. Furthermore, most existing automation solutions rely on finite element simulation for offline compensation calculations, which cannot meet the real-time requirements of production lines.

[0004] Furthermore, single-sensor sensing methods suffer from poor stability in complex environments and lack robustness against multiple interference sources. Multi-gripper systems also generally lack effective collaborative control mechanisms, making path conflicts and motion interference more likely to occur, impacting system stability.

[0005] Therefore, it is necessary to develop an intelligent fixture adjustment system with high-precision sensing capability, rapid compensation mechanism and multi-fixture coordination capability to improve the efficiency and quality of flexible assembly. Summary of the Invention

[0006] In view of this, in order to address the problems of poor adaptability, slow response speed, unstable perception and multi-fixture adjustment conflicts in the existing fixture system, this paper provides an adaptive fixture adjustment method based on intelligent vision and pneumatic control to achieve high-precision, multi-modal perception and real-time fixture compensation control.

[0007] In order to solve the above problems, the technical solutions adopted by the present invention are as follows:

[0008] An adaptive fixture adjustment method based on intelligent vision and pneumatic control, comprising:

[0009] Multimodal perception module, used to collect three-dimensional deformation information of parts, integrating binocular vision, strain sensors and laser displacement meters to achieve full-domain multi-dimensional perception;

[0010] A hybrid modeling module combines the stiffness matrix constructed by finite element method with the XGBoost machine learning model to calculate the real-time compensation of the fixture;

[0011] Dynamic optimization module, using PSO-AdaptivePenalty algorithm and fuzzy PID control strategy to optimize regulation accuracy, energy consumption and action safety;

[0012] The pneumatic actuator module achieves sub-millimeter displacement adjustment through a double-acting cylinder combined with a cam-groove mechanism, and provides closed-loop feedback of the clamping force through a sensor.

[0013] The beneficial effects of the present invention are:

[0014] The system can achieve real-time fixture adjustment with high precision (error controlled within ±0.002mm) and high response speed (control cycle does not exceed 200ms) in complex assembly environments. It supports collaborative operation of multiple fixtures, has good scalability and robustness, and is suitable for high-end intelligent manufacturing fields such as automobiles and aerospace. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is the overall structural block diagram of the method of the present invention;

[0016] Figure 2 Schematic diagram of the working principle of the multimodal perception module in the present invention;

[0017] Figure 3 The calculation flow chart of the hybrid modeling module;

[0018] Figure 4 Schematic diagram of the multi-objective adjustment process in the dynamic optimization module;

[0019] Figure 5 This is a structural cross-sectional view of the pneumatic actuator module. DETAILED DESCRIPTION

[0020] In order to make the objectives, technical solutions and advantages of the present invention more clear, the present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0021] The present invention proposes an adaptive fixture adjustment method based on intelligent vision and pneumatic control, which is aimed at the clamping needs of special-shaped, deformed or flexible parts in current industrial assembly, and has high-precision perception, rapid response adjustment and multi-fixture coordination capabilities.

[0022] The method adopts a technical solution that combines multimodal sensing integration, hybrid model construction and optimization control algorithm, which significantly improves the adaptability and stability of fixture adjustment under flexible manufacturing conditions.

[0023] like Figure 1 As shown in the figure, this method constructs a four-layer control architecture of "multimodal perception-hybrid modeling-dynamic optimization-pneumatic execution".

[0024] A closed-loop logic chain is formed between each functional module to achieve full-process response from perception to modeling to decision execution.

[0025] The method is suitable for high-complexity scenarios where multiple fixtures work in parallel, and can be widely used in automotive chassis parts, aviation skin components, consumer electronics mid-frames and other fields.

[0026] like Figure 2 As shown in Figure 1, the perception module integrates the following three types of sensors:

[0027] The binocular vision submodule uses dual cameras to build a point cloud model, aligns the CAD reference model using the improved ICP algorithm, and reconstructs the workpiece contour;

[0028] Fiber optic strain sensors, placed near the contact edge of the fixture, can obtain subtle local tension / compression responses;

[0029] Laser displacement sensor for non-contact precise distance measurement, combined with vision compensation for highly reflective / highly obscured areas.

[0030] After the three types of data are synchronized in time and calibrated in the spatial domain, the system forms the workpiece deformation field vector:

[0031] ΔD=[d1,d2,...,d n ] T

[0032] where d i It represents the displacement deviation of the i-th key feature point on the workpiece, in mm, and the overall resolution is controlled within ±0.01mm.

[0033] like Figure 3 As shown in the figure, the control flow of the fixture adjustment method is as follows:

[0034] Collect 3D point clouds through binocular cameras to extract the geometric features of the workpiece;

[0035] Use SIFT algorithm to identify local deformation areas and calibrate the target pose in the clamping area;

[0036] According to the deformation vector ΔD, the compensation amount ΔP is generated by the hybrid modeling module;

[0037] Control the translation, rotation and other degrees of freedom of the fixture base to achieve precise adjustment;

[0038] Start the pneumatic system, and the double-acting cylinder applies the clamping force to complete the clamping;

[0039] The end sensor detects the clamping status and recalibrates or updates the modeling parameters if the error exceeds the threshold.

[0040] Among them, the fixture compensation amount ΔP represents the displacement vector of each adjustment degree of freedom and can be expressed as:

[0041] ΔP=[x,y,z,θ x ,θ y ,θz ] T

[0042] It represents the six degrees of freedom adjustment of the fixture in three-dimensional translation and rotation around three axes, and the hybrid modeling module will generate an accurate solution later.

[0043] The above adjustment strategy enables the fixture to adapt to changes in the state of complex workpieces through a full-process perception-modeling-execution coupling mechanism, avoiding alignment errors caused by manual adjustment or fixed molds, and significantly improving the degree of automation and production efficiency.

[0044] After obtaining the deformation field vector ΔD of the workpiece, the system enters the hybrid modeling and dynamic optimization stage.

[0045] This part includes key steps such as stiffness modeling, residual compensation, optimization calculation and control adjustment, aiming to generate the fixture adjustment amount ΔP that meets the accuracy, energy consumption and safety constraints.

[0046] The hybrid modeling module of the present invention comprehensively adopts physical models and machine learning models to achieve a fixture adjustment modeling method that is highly accurate, explainable, and adaptable.

[0047] The physical modeling part is based on finite element simulation. In the initial sample stage, the global stiffness matrix K of the workpiece is calculated through ANSYS or MATLAB simulation, and a linear relationship between the input compensation ΔP and the output deformation ΔD is established:

[0048] ΔD=K·ΔP

[0049] However, considering the existence of nonlinear deformation of materials and systematic errors in actual working conditions, the linear model has residual deviations.

[0050] To this end, the present invention introduces a machine learning model - incremental XGBoost regressor to perform online prediction of residual terms to compensate for the shortcomings of the physical model.

[0051] Let the prediction residual be ΔP residual , then the final adjustment expression is:

[0052] ΔP=K + ΔD+ΔP residual

[0053] where K + It represents the Moore-Penrose pseudo-inverse of the stiffness matrix, which is convenient for solving problems with redundant degrees of freedom or ill-conditioned problems.

[0054] ΔP residual Training is driven by historical adjustment samples and real-time feedback errors. The regression model parameters are automatically updated through each round of adjustment results to achieve real-time self-adaptation.

[0055] like Figure 4 As shown in Figure 2, after obtaining the adjustment value ΔP, the system further uses the dynamic optimization module to solve the constraints of the multi-objective adjustment task. The objectives include:

[0056] Minimize the total displacement error ||ΔD-K·ΔP|| 2 ;

[0057] Control system energy consumption, i.e., minimize the amount of actuator kinetic energy or air pressure used;

[0058] Ensure that there is no physical interference or collision in the adjustment paths between the fixtures.

[0059] The present invention uses the particle swarm optimization algorithm and combines it with the adaptive penalty factor strategy to construct a comprehensive objective function:

[0060] min J(ΔP)=α1·E deform +α2·E energy +α3·E collision

[0061] Among them, E deform Denotes deformation error, E energy Indicates execution energy consumption estimation, E collision represents the path conflict risk between fixtures; α i is the weight coefficient that is dynamically adjusted according to the real-time working conditions.

[0062] In addition, in order to improve the smoothness and anti-interference ability of the regulation response, the system introduces a fuzzy PID control mechanism.

[0063] The controller uses deformation error e=||ΔD 目标 -ΔD 实际 || and error change rate e c As input, fuzzy inference rules are used to dynamically adjust K p , K i , K d Parameters are used to achieve adaptive matching of nonlinear response characteristics. The specific control law is as follows:

[0064]

[0065] Where u(t) is the adjustment command signal, which is executed by the pneumatic control module.

[0066] The dynamic optimization module also integrates a distributed arbitration protocol, supporting up to 16 fixtures working in parallel. Through timestamp synchronization and path planning (based on the RRT* algorithm), it detects conflicts and arranges the sequence of fixture movements, avoiding intersections or occlusions in the gripping paths.

[0067] Through the above modeling and optimization process, this method can achieve high-precision, fast, and safe adaptive fixture adjustment in complex clamping scenarios, and has good real-time and generalization capabilities.

[0068] After the adjustment compensation amount ΔP is output by the optimization module, the system converts the numerical instruction into actual mechanical displacement through the pneumatic execution module, thereby completing the precise control and clamping action of the fixture.

[0069] like Figure 5 As shown, the pneumatic actuator module (4) consists of the following main components:

[0070] A double-acting cylinder (4a) for providing a controllable and reversible linear driving force;

[0071] The cam-groove mechanism (4b) converts linear input into a small displacement output with nonlinear buffering characteristics, which is suitable for the high-precision adjustment requirements of the fixture;

[0072] A pressure sensor (4c) monitors the reaction force generated during the clamping process in real time to prevent overpressure or clamping failure;

[0073] The control feedback interface (5) forms a closed-loop execution and monitoring with the control module.

[0074] This module maps the information of each degree of freedom in the optimized target compensation amount ΔP, drives each adjustment unit of the fixture to complete precise movement, and the minimum control accuracy of the adjustment range can reach ±0.02mm.

[0075] The control method supports displacement-force decoupling, that is, the actuator can independently adjust the fixture position and clamping force during the adjustment process, avoiding damage to flexible or easily deformed parts.

[0076] In order to ensure the stability and predictability of the adjustment process, the present invention also introduces a digital twin verification module.

[0077] The system establishes a simulation model of the real actuator in the MATLAB Simulink platform, covering the aerodynamic response characteristics, structural flexibility characteristics and typical disturbance conditions.

[0078] Before execution, the convergence and response time of the adjustment instructions can be verified through virtual trial operation to form a closed-loop adjustment strategy.

[0079] After each round of adjustment, the system compares the expected deformation field with the actual perception result and calculates the residual term ε:

[0080] ε=||ΔD 实际 -ΔD 目标 ||

[0081] If the residual exceeds the system set threshold ε thresh, the system will trigger the self-learning mechanism, perform incremental training on the XGBoost residual model, and automatically update the modeling parameters, thereby further improving the subsequent adjustment accuracy.

[0082] In addition, the present invention combines finite element thermal simulation to assist in verifying the force concentration and deformation distribution of the fixture structure.

[0083] like Figure 5 As shown on the right, the simulation results show the stress / heat concentration areas of the adjustment fixture under different loads, providing a basis for optimizing the structural design and control strategy.

[0084] This embodiment implements intelligent fixture adjustment in highly complex environments through the complete perception-modeling-optimization-execution closed-loop process described above, and has the following comprehensive advantages:

[0085] The precision is significantly improved: the fixture clamping error is reduced from the traditional ±0.5mm to ±0.002mm;

[0086] Shortened response time: compensation control cycle is controlled within 200ms;

[0087] Strong multi-fixture scalability: supports parallel and coordinated adjustment of up to 16 fixtures to avoid conflicts and interference;

[0088] Stable and reliable control: Combining fuzzy control with physical / learning models enables rapid adjustment and self-correction.

[0089] The present invention has broad application prospects in intelligent manufacturing scenarios and is particularly suitable for online high-precision clamping operations of complex parts in flexible production lines.

[0090] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention's description and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. An adaptive fixture adjustment method based on intelligent vision and pneumatic control, characterized in that: It includes: Multimodal perception module, used to obtain spatial deformation information of the measured part, including binocular stereo vision submodule, fiber optic strain sensor and laser displacement meter; Hybrid modeling module, used to construct the stiffness matrix based on finite elements and the XGBoost model to predict the residual compensation amount and output the fixture adjustment vector ΔP; The dynamic optimization module uses particle swarm optimization and fuzzy PID joint control strategy to perform multi-objective optimization of the compensation amount; The pneumatic actuator module achieves sub-millimeter clamp displacement adjustment through a double-acting cylinder and a cam-groove transmission mechanism.

2. The method according to claim 1, characterized in that The multimodal perception module uses SIFT feature matching and adaptive threshold segmentation algorithms to process visual data, achieving sub-pixel edge extraction capabilities under illumination of 300 to 1000 lux.

3. The method according to claim 1, characterized in that The hybrid modeling module integrates physical modeling and data-driven modeling, where: The finite element physical model generates the stiffness matrix K and establishes the deformation-compensation relationship as follows: ΔD=K·ΔP XGBoost model is used to predict nonlinear residual compensation ΔP residual , the final compensation amount is: ΔP=K + ·ΔD+ΔP residual Among them, K + represents the pseudoinverse of the stiffness matrix.

4. The method according to claim 1, wherein The dynamic optimization module adopts the PSO-AdaptivePenalty algorithm to simultaneously optimize the following three objective functions: fixture deformation compensation accuracy, system energy consumption minimization, and collision-free constraints between fixtures.

5. The method according to claim 1, wherein The fuzzy PID controller is based on the error e and the error change rate e. c As input, adaptively adjust the gain parameter K p , K i , K d , the overshoot of the control system adjustment response does not exceed 5%.

6. The method according to claim 1, characterized in that The pneumatic actuator module uses a double-acting cylinder combined with a cam-groove structure to convert air pressure into linear displacement with a displacement accuracy of ±0.02mm. The clamping force is fed back through a closed-loop pressure sensor to achieve force-displacement decoupling control.

7. The method according to claim 1, characterized in that The system supports dynamic collaborative adjustment of up to 16 fixtures, and achieves zero-collision path planning between fixtures through a distributed priority arbitration protocol and the RRT* path planning algorithm.

8. The method according to claim 1, characterized in that This method builds a digital twin model in MATLAB Simulink for virtual regulation simulation and control strategy convergence verification. When the residual: e=||ΔD 实际 -ΔD 目标 || When the set threshold is exceeded, the system triggers an incremental update of the XGBoost model.

Images