An additive forming connection area local in-situ flexible processing device and processing method

Abstract

A kind of additive forming connecting area local in-situ flexible processing device and processing method belong to intelligent manufacturing and robot processing equipment field.The processing device includes omni-directional mobile platform, industrial robot, end effector group, double compliant force control polishing head and control system.Processing method is that omni-directional mobile platform drives to workpiece station, quick-change disc switches three-dimensional scanner to complete hand-eye calibration online verification, automatically scans additive forming connecting area, two-stage registration generates double-channel allowance distribution matrix, adaptive tool path generation and simulation verification, quick-change disc switches milling spindle to carry out double pre-compensation rough milling, intermediate secondary scanning correction finishing tool path, finish milling, quick-change disc switches double compliant force control polishing head and executes partition force control polishing, final inspection scanning and determines eligibility.Can realize " drive to, scan, milling, polishing " in-situ processing in workpiece parking station, avoid large frame structure turnover transportation, improve additive forming connecting area processing precision and surface quality consistency.

Description

Technical Field

[0001] This invention belongs to the field of intelligent manufacturing and robotic processing equipment technology, specifically relating to a local in-situ flexible processing device and processing method for additive forming connection areas. Background Technology

[0002] Aircraft frame structures (such as fuselage frames, wing ribs, and door frames) are typically made of aluminum alloy. To shorten the manufacturing cycle of large integral components and reduce material removal, a hybrid manufacturing model of "additive forming + forming connection" has been widely adopted in recent years. Material is deposited at the joining area of ​​the machined base material to form the additive forming connection area, which is then further processed in situ to achieve the final dimensions and surface quality requirements.

[0003] Compared with traditional blank processing, additive forming of the connection zone has the following three prominent features: (1) The allowance distribution is extremely uneven. Due to the influence of deposition path, process parameter fluctuations, interlayer temperature unevenness and thermal deformation, the allowance deviation of the measured surface of the additive forming connection area relative to the theoretical digital model can reach ±1.5~5mm, and the difference is significant at different positions on the same surface.

[0004] (2) Significant non-rigid deformation. The thermal cycle of additive deposition creates a residual stress field inside the additive forming connection area, resulting in local warping and twisting of the additive forming connection area. In addition to the overall rigid body offset, the deviation between the measured point cloud and the theoretical digital model also includes local shape changes of 0.5~2mm, i.e., non-rigid deformation.

[0005] (3) The material properties are significantly different. There are differences in hardness and work hardening behavior between the additive manufacturing region and the base material region (e.g., 5-series aluminum alloys 5052, 5083, 5A06). The interface transition zone is gradient-distributed, resulting in significant differences in cutting force-deformation response and grinding removal rate in different regions.

[0006] The processing of the additive manufacturing connection area of ​​existing large frame structures mainly adopts the following four technical solutions, but all of them have varying degrees of shortcomings.

[0007] The first method involves machining on a large gantry milling machine. During machining, the additively manufactured components need to be transported to a large gantry CNC machine tool, clamped, and then machined offline based on a nominal digital model. This method has the following drawbacks: The transportation, hoisting, and alignment of large frame structures within the workshop are time-consuming, typically accounting for more than 30% of the entire process cycle, and there is a risk of damage during transportation; the fixed toolpath based on the nominal digital model cannot accommodate allowance deviations of ±1.5~5mm, easily leading to situations where "where there is too much allowance, it is not cut enough; where there is too little allowance, it is cut too much"; and it cannot pre-compensate for delayed deformation caused by residual stress release after milling, often requiring multiple rework and grinding operations.

[0008] The second type is the existing mobile robot machining system. For example, CN109746928A discloses a mobile dual-robot automatic milling flexible production line, and CN117841012A discloses a method for constructing the milling error dimension chain of a mobile robot, which integrates industrial robots into a mobile platform to achieve in-situ machining. However, the above solutions mainly address the problems of mobile positioning and dimension chain. Their registration process only uses six-degree-of-freedom rigid body iterative nearest point (ICP) fitting, which cannot absorb the non-rigid deformation unique to the additive forming connection area, causing the registration residual to be directly included in the allowance calculation.

[0009] The third approach is the existing adaptive machining method. For example, CN112710228A discloses a robot milling path planning method based on global measurement, and CN202610338659 discloses a robot layered grinding method, both employing a strategy of "scanning → Boolean intersection of measured and theoretical digital models → partitioning by allowance." While these schemes perform coarse and fine partitioning based on allowance size, the allowance grading is based on a single material assumption and does not consider the performance gradient at the additive / base material interface; furthermore, they do not establish a two-dimensional parameter selection mechanism of "geometric allowance × material partitioning" for the additive forming connection zone. In addition, existing stiffness compensation patents (such as CN112893955B and CN111007804A) only compensate for instantaneous elastic deformation induced by cutting force, failing to address the delayed deformation caused by residual stress release after milling the additive forming connection zone.

[0010] The fourth type is the existing robotic force-controlled grinding system. For example, the passive compliance and active compliance hybrid control method disclosed in CN106041933A achieves grinding force control through a six-dimensional force sensor and a passive compliance device. However, in the above scheme, the normal force setting value is constant, and there is no adaptive switching of force control parameters for material property gradient regions; it also lacks an online interface identification method that combines the prior position of material partitions with the force signal spectrum characteristics, which can easily cause sudden changes in force and surface quality when the grinding wheel crosses the additive / base material interface.

[0011] In summary, none of the existing technologies have provided a systematic solution to the three unique problems of additive forming connection areas: "extremely uneven allowance distribution, significant non-rigid deformation, and obvious material property gradient". Summary of the Invention

[0012] To address the problems existing in the prior art, the present invention aims to provide a local in-situ flexible processing device and method for additive manufacturing connection areas, so as to solve the following technical problems: (1) There is non-rigid deformation between the measured point cloud and the theoretical model in the additive forming connection area. The traditional rigid body ICP registration residual pollution margin calculation leads to inaccurate margin calculation and incorrect toolpath cutting. (2) After milling to remove the allowance, the residual stress inside the additive forming connection area is rebalanced and released, which causes the additive forming connection area to continue to deform for several hours after the machining is completed. The dimensions that were qualified during machining are out of tolerance afterward. The existing stiffness compensation method can only compensate for the problem of instantaneous elastic deformation induced by cutting force. (3) The material property gradient between the additive zone and the base material zone leads to inconsistent grinding removal rates on both sides of the interface under constant force control parameters, and the grinding wheel generates instantaneous force change impact across the interface. (4) Large frame structure components need to be transported to machine tools for processing, which results in long clamping time and low efficiency.

[0013] To achieve the above objectives, this invention provides a local in-situ flexible machining device and method for additive forming connection areas. The local in-situ flexible machining device for additive forming connection areas adopts an in-situ machining mode of "equipment movement replacing workpiece movement": a manually remotely controlled omnidirectional moving platform carries an industrial robot to the workpiece parking position. After the four hydraulic outriggers lift and unload the load from the traveling wheels, the industrial robot sequentially switches between a 3D scanner, a milling spindle, and dual compliant force-controlled grinding heads via an end-effector pneumatic quick-change disc. The machining of the additive forming connection area is completed according to a closed-loop process chain of "scanning → two-level registration → dual-channel allowance distribution matrix → adaptive toolpath → dual pre-compensation rough milling → intermediate secondary scan correction → finish milling → zoned force-controlled grinding," particularly suitable for machining large aluminum alloy components such as aircraft frame structures.

[0014] The additive forming connection region to which this invention applies includes: a local area deposited on a base material substrate by laser melting deposition, arc additive manufacturing, or similar additive manufacturing processes for repair, supplementation, or connection, and a transition region in contact with the base material. Due to the residual stress field generated by the thermal cycle of additive deposition, the additive forming connection region exhibits non-rigid warping deformation relative to the theoretical CAD model, in addition to rigid body offset. Simultaneously, a material property gradient exists between the additive region and the base material region, posing unique challenges to the accuracy and surface quality consistency of subsequent milling and grinding.

[0015] To achieve the above objectives, the technical solution adopted by the present invention is as follows: The present invention provides a local in-situ flexible processing device for the bonding area of ​​additive forming, comprising an omnidirectional moving platform, an industrial robot, an end effector assembly, a dual compliant force control grinding head, and a control system; An industrial robot is mounted on the omnidirectional mobile platform to control its movement to the workpiece parking position. The end effector of the industrial robot is connected to an end effector assembly, which automatically switches between three end effectsors via a quick-change disc: a 3D scanner, a milling spindle, and a dual compliant force-controlled grinding head. The dual compliant force-controlled grinding head consists of a six-dimensional force sensor, a pneumatic floating mechanism, and an eccentric vibrating grinding disc connected in series along the axial direction. The control system is used to automate the local in-situ flexible processing device in the additive forming connection area.

[0016] The omnidirectional mobile platform adopts a Mecanum wheel omnidirectional mobile mechanism, is equipped with a manual remote control operating system, and has four hydraulic outriggers. During processing, the four hydraulic outriggers are raised to unload the load from the traveling wheels; the load-bearing capacity is not less than 3.5 tons.

[0017] The industrial robot has a rated load of 500kg, a maximum working radius of 2826mm, and a repeatability of ±0.08mm.

[0018] The quick-change disc is preferably a pneumatic quick-change disc with a repeatability of ≤ ±0.02mm, a measurement accuracy of 3D scanner of ≤ 0.02~0.03mm, a rated power of milling spindle of ≤ 15kW, S1 continuous duty cycle (S1 continuous duty cycle refers to the motor being able to run continuously for a long time under rated load), a maximum speed of ≥ 24000rpm, HSK-E50 tool interface, and cooling methods including water cooling, oil cooling, and MQL (Minimum Quantity Lubrication).

[0019] The six-dimensional force sensor has a sampling frequency of no less than 1000Hz, and the pneumatic floating mechanism has a floating stroke of ±5~8mm and a preload pressure adjustable within the range of 0.2~0.6MPa. An impedance controller is used to control the entire dual-compliant force-controlled grinding head, with an impedance control cycle of 1kHz. The pneumatic floating mechanism includes a cylinder, a linear guide, and a preload spring installed in series. The cylinder preload pressure is adjusted in real time via a proportional pressure regulating valve according to the corresponding value in the partitioned force control parameter table.

[0020] The control system includes a two-level registration and allowance matrix module, an adaptive toolpath generation module, a stress release + stiffness dual pre-compensation module, a zoned force control and safety monitoring module, and a digital twin simulation verification module. The two-level registration and allowance matrix module is used to obtain and provide feedback on the machining allowance. The adaptive toolpath generation module is used to determine and provide feedback on the robot's machining trajectory. The stress release + stiffness dual pre-compensation module is used to provide part accuracy compensation. The zoned force control and safety monitoring module is used to provide safe and compliant grinding. The digital twin simulation verification module is used to provide virtual simulation of the entire machining process.

[0021] The control system also includes a hand-eye calibration and verification module, which is used to scan and check the overlap of the center position of a miniature reference ball fixed to the mobile platform surface in 3 to 5 postures after each start-up or replacement of the 3D scanner. When the standard deviation of the center position exceeds 0.05mm, a recalibration prompt is given.

[0022] The additive forming connection area local in-situ flexible processing device also includes a safety monitoring module. When the normal contact force exceeds the hard upper limit of 150N, the safety monitoring module triggers an emergency stop and drives the industrial robot to retreat along the normal direction.

[0023] The present invention provides a method for local in-situ flexible processing of the additive forming connection area, comprising the following steps: The omnidirectional moving platform moves to the workpiece station, the four hydraulic outriggers lift it up, the quick-change disc of the end effector group switches to the 3D scanner and completes online verification of hand-eye calibration, automatically scans the additive forming connection area, generates a dual-channel allowance distribution matrix through two-level registration, generates and simulates the adaptive toolpath, the quick-change disc switches to the milling spindle and performs dual pre-compensation rough milling, intermediate secondary scanning and correction of the finishing toolpath, finish milling, the quick-change disc switches to dual compliant force-controlled grinding heads and performs zoned force-controlled grinding, and finally scans and determines the pass rate.

[0024] Furthermore, the quick-change disc switching of the 3D scanner and the completion of online hand-eye calibration verification are achieved by using the measured point cloud obtained from the 3D scanner held by the end effector of an industrial robot and performing preprocessing: The measured point cloud scanned by the 3D scanner is denoised and simplified, and then meshed to reconstruct a triangular mesh surface. Using the tooling positioning surface and the machined surface of the base material as reference features, the meshed measured point cloud and the theoretical CAD model are subjected to constrained weighted rigid body iterative nearest point (ICP) registration, and the six-degree-of-freedom rigid body transformation is solved. In each iteration, the corresponding point pairs between the measured points and the theoretical model are established through nearest neighbor search. After removing outliers with a distance greater than 2mm, the weighted registration objective function is constructed as follows:

[0025] Where m is the total number of point pairs participating in registration, w i Let x be the matching weight for this point pair, R be a 3×3 rotation matrix, and x be the value of the matching weight. i Let be the i-th point in the measured point cloud, t be the three-dimensional translation vector, and y be the i-th point. i For theoretical numerical models and x i The nearest corresponding point; w of each point in the reference feature point set iThe matching weight of each point in the additive forming connection zone is taken as 5 to 10 times; in each iteration, singular value decomposition (SVD) is used to find the optimal rigid body transformation that minimizes the objective function, with the rotation angle change between adjacent iterations being less than 10°. -5 rad and the translation change is less than 10 -4 mm is used as the convergence condition; based on the rigid body registration results, the TPS deformation field is fitted to the additive forming connection area, and the optimization objective is the weighted sum of the registration residual and the bending energy. The reference features for the constraint-weighted rigid body iterative nearest point (ICP) registration include at least one of the following: tooling positioning surface, machined surface of the base material, fixture reference pin hole, and platform table reference point calibrated by a coordinate measuring machine.

[0026] The TPS deformation field fitting for the additive forming connection region is based on the rigid body registration results, by superimposing a non-rigid deformation field f(x) of a thin plate spline (TPS) based on radial basis functions (RBF) onto the point cloud of the additive forming connection region. The deformation function is: Where x is the three-dimensional position vector of the surface point of the additive forming connection area (in the robot base coordinate system, the same below). This is the radial basis function of the TPS, where the independent variable r = ||xc|| represents the Euclidean distance from a spatial point x to a control point c, and c... i Let w be the i-th control point, i = 1, 2, ..., n, where n is the total number of control points. i The weights are to be determined and then processed by regularization. The bending energy is controlled, where λ is the regularization weighting coefficient used to balance the relative weight of the registration residual term and the bending energy term. The optimization objective is to minimize the remaining registration residual and the bending energy simultaneously, and then solve for the deformation field coefficients.

[0027] Further calculation of the alignment coordinates x′ of the measured points after two-stage registration in the theoretical numerical model coordinate system yields the following transformation model:

[0028] Where x′ is the aligned coordinate of point x in the measured point cloud after the two-stage registration in the theoretical numerical model coordinate system, that is, the position of the point in the aligned measured point cloud; the regularization term is taken as the TPS bending energy. , It is a second-order partial derivative operator that ensures f(x) absorbs only the local warping deformation of the additive forming connection region without disrupting the global reference relationship established by rigid body registration. The undetermined weight w of the TPS deformation field. iThe solution is obtained directly by taking the closed solution of the linear equation system formed by the corresponding points, the residual term and the regularization term; at the same time, the anchoring constraint of f(x)=0 is applied to all reference feature points to ensure that the deformation field does not affect the aligned reference area.

[0029] The two-level registration generating dual-channel margin distribution matrix refers to aligning the measured point cloud to the theoretical digital model according to the two-level registration model after registration, and generating a dual-channel margin distribution matrix: the first channel is the geometric margin value h(x) at each tool point, defined as the signed distance from the theoretical surface to the measured surface along the tool axis direction (i.e., the normal direction of the machining surface), and takes a positive value when the measured surface is outside the theoretical surface; the second channel is the material partition identifier at the corresponding tool point, including the additive manufacturing area, the transition area and the base material area.

[0030] Furthermore, the material zoning identifier is determined based on the material zoning map. The method for generating the material zoning map is as follows: the outline of the additive manufacturing area is taken from the deposition boundary given by the additive deposition path planning data; the transition zone is the ring formed by extending the outline of the additive manufacturing area outward by 5~10mm along the surface normal; the remaining areas are identified as the parent material area; after the two-level registration is completed, the material zoning map is superimposed on the measured point cloud in the same coordinate system, and the zoning to which it belongs is queried for each cutter site.

[0031] The adaptive toolpath generation and simulation verification involves selecting cutting parameter combinations by looking up a table based on a two-dimensional "allowance value × material partition" during toolpath generation to achieve adaptive toolpath generation.

[0032] The rule for selecting cutting parameters based on a two-dimensional lookup table using "allowance value × material zone" is as follows: when the geometric allowance h > h 精 When the material is in the additive manufacturing zone or transition zone, rough machining is performed with a depth of cut ap = 2.0 mm, a width of cut ae = 10 mm, feed rate F = 2300 mm / min, and cutting speed Vc = 450 m / min; when 0 <h≤h 精 When performing finishing, ap ≤ 0.5 mm, F ≤ 1000 mm / min; when h ≤ 0, it is marked as empty cutting and the cutting point is skipped; the finishing allowance threshold h 精 Take 0.5mm.

[0033] After the adaptive toolpath is generated, collision detection and digital twin simulation are performed on the robot to verify that there is no collision and the predicted end-effector deformation does not exceed 0.1mm before milling is performed.

[0034] Furthermore, to address the issue of delayed deformation in the additive forming connection area after material removal, this invention performs reverse compensation by superimposing the stress release deformation δ1(x) and the cutting force-induced elastic deformation δ2(x) vectors, and achieves actual measurement correction through intermediate secondary scanning.

[0035] The method for predicting stress release deformation δ1(x) is as follows: Based on the dual-channel residual distribution matrix obtained by scanning, a local finite element model of the additive forming connection area is established. The initial residual stress field in the finite element model is assigned according to the simulation results of additive manufacturing process or the measured residual stress data (the measured data can be obtained by X-ray diffraction or drilling method). The removed elements are gradually deleted according to the layered removal sequence of the roughing toolpath, and the deformation field caused by the rebalancing of residual stress in each processing stage is simulated and predicted.

[0036] The prediction method for the elastic deformation δ2(x) of the cutting force is as follows: Establish a static stiffness model of the entire machine, including the stiffness of the hydraulic outriggers of the mobile platform, the stiffness of the robot joints, and the stiffness of the linkage structure; Then, input the robot joint angle vector q corresponding to the current tool position and the end-effector cutting force vector F. c Substituting the Cartesian stiffness matrix K(q) at the end of the machine, the elastic deformation is estimated. Where K(q) is composed of the robot Jacobian matrix J(q) and the joint stiffness diagonal matrix K. θ According to the mapping relationship K(q)=J - ᵀ·K θ ·J -1 The calculation yields J, where J T It is the transpose of the Jacobian matrix, K θ J is the joint stiffness diagonal matrix. -1 It is the inverse of the Jacobian matrix; then it is synthesized with the stiffness of the hydraulic outrigger and the stiffness of the linkage structure according to the series and parallel spring model; the stiffness of the linkage structure is obtained through finite element analysis, and the stiffness of the robot joint is obtained through technical specifications or stiffness identification experiments; the cutting force vector is obtained based on the empirical cutting force model or finite element cutting simulation established by the tool diameter, cutting depth, cutting width, feed rate and cutting speed.

[0037] In the toolpath post-processing stage, δ1(x) and δ2(x) are first uniformly transformed to the robot base coordinate system for vector superposition. For each tool position point, the component of the comprehensive deformation prediction value δ1+δ2 in the normal direction of the machining surface is taken, and the tool position point coordinates are shifted in the opposite direction of this component so that the actual surface after machining falls on the theoretical surface. After roughing and before finishing, after rough milling, the surface deformation increment is allowed to stand until it is less than the set threshold in two consecutive measurements. Then, an intermediate secondary scan is performed to rescan the rough milled surface and measure the actual residual deviation after stress release. When the measured residual deviation is greater than the set threshold, the finishing toolpath is corrected according to the measured deviation. After completing the intermediate secondary scan and finishing toolpath correction, the milling spindle is switched again through the pneumatic quick-change disc to perform finishing milling with a cutting depth of ap≤0.5 mm. When the residual deviation is greater than the set threshold (typically 0.05 mm), the finishing toolpath is corrected according to the measured deviation, forming a "prediction + measurement correction" double closed loop.

[0038] The intermediate secondary scan uses the same scanner and hand-eye calibration matrix as the pre-rough milling scan to ensure consistent measurement reference.

[0039] The quick-change disc switching dual compliant force-controlled grinding head and the execution of zoned force-controlled grinding address the issue of interface material performance gradient. The dual compliant force-controlled grinding head of this invention adopts an architecture that couples active impedance control with passive pneumatic floating, and introduces a material zoned force control parameter lookup table switching mechanism.

[0040] Active impedance control includes a six-dimensional force sensor and an impedance controller. The impedance controller is used to determine the deviation between the normal contact force measured by the six-dimensional force sensor and the target normal force in the partitioned force control parameter table. The six-dimensional force sensor measures the normal contact force in real time, and the impedance controller operates according to a second-order impedance model with a control cycle of 1 kHz. The normal force deviation is mapped to a position correction Δx and superimposed on the original grinding trajectory, where M is the virtual inertia, B is the damping, and K is the position correction value. s For virtual stiffness, F n,des For the target normal force in the force control parameter table of the partition, F n,meas The normal contact force was measured using a six-dimensional force sensor. The second derivative of the position correction (acceleration). The first derivative of the position correction (velocity). This value is constantly changing and serves as the position correction amount. The difference from traditional constant-force grinding is that the normal force setting and impedance parameters are no longer constant, but rather switch according to the pre-set material zone force control parameter table. The additive manufacturing zone, transition zone, and base material zone correspond to different normal force settings, impedance parameters, and pneumatic preload pressures. The zone force control parameter table stores the normal force setting, virtual inertia M, damping B, and virtual stiffness K for the additive manufacturing zone, transition zone, and base material zone, respectively. s And pneumatic preload pressure. The normal force setting value is 60~80N in the additive manufacturing zone, 50~65N in the transition zone, and 45~60N in the base material zone; the virtual stiffness K s The transition zone is lower than the base material zone to buffer the sudden force change when the grinding wheel crosses the interface.

[0041] The passive pneumatic float is a pneumatic floating mechanism connected in series between the six-dimensional force sensor and the eccentric vibrating grinding disc. The floating stroke is ±5~8mm, and the cylinder preload pressure switches synchronously with the material partition within the range of 0.2~0.6MPa. The active impedance control focuses on low-frequency large deviation compensation, while the passive pneumatic float absorbs high-frequency force disturbances outside the bandwidth of the active impedance control. In the event of a failure in the active impedance control, the passive pneumatic float can still maintain an acceptable contact force, providing a safety redundancy.

[0042] The final inspection scan is an interface recognition mechanism employing a dual-channel redundancy system: Channel 1 scans the material partition map to identify the additive / base material interface online based on the prior interface position and / or the spectral characteristics of the normal force signal, triggering parameter switching along the grinding trajectory; Channel 2 detects the spectral characteristics of the normal force signal in the 50-200Hz frequency band online, calculating the spectral energy using a 0.5-1s sliding time window. When the abrupt change exceeds 20%, it is determined that the grinding wheel has crossed the interface. When the grinding wheel crosses the interface, the force fluctuation spectral characteristics change abruptly. After real-time identification, it is cross-checked with the prior position, triggering the impedance controller and pneumatic floating mechanism to switch to the force control parameters of the corresponding partition. Triggering any channel automatically switches the force control parameter group in advance to avoid sudden force shocks. The safety monitoring module triggers an emergency stop and drives the industrial robot to retreat along the normal direction when the normal force exceeds the 150N hard upper limit.

[0043] This invention addresses three unique problems in additive forming connection areas: non-rigid warping deformation caused by thermal residual stress, delayed deformation after residual stress release, and material property gradient at the additive / parent material interface. The following methods are proposed: (1) Two-level registration combining constrained weighted rigid body ICP and RBF thin plate spline (TPS) non-rigid deformation field to generate a dual-channel allowance distribution matrix of geometric allowance and material partition, and generate adaptive toolpath by looking up the table according to the allowance value × material partition in two dimensions; (2) Superimpose the stress release deformation predicted by finite element layer removal with the cutting force elastic deformation vector predicted by the overall machine stiffness matrix to pre-compensate the tool position in reverse, and insert intermediate secondary scanning after rough milling to measure and correct the finishing toolpath; (3) Use a dual compliant force control grinding head coupled with active impedance control and passive pneumatic floating, switch the normal force setting and impedance parameters by looking up the table according to the material partition, and identify the interface online through the dual-channel dual-channel of the material partition a priori position and force signal spectrum characteristics.

[0044] Compared with the prior art, the present invention has the following beneficial effects: (1) Eliminate the pollution of non-rigid deformation on the allowance calculation. Compared with single rigid body ICP, two-level registration (constrained weighted rigid body ICP + RBF / TPS non-rigid deformation field) can reduce the registration residual of the additive forming connection area from the millimeter level to the order of 0.05 mm level, and the accuracy of allowance calculation is improved by an order of magnitude.

[0045] (2) Solve the problem of post-processing deviation caused by stress release delay deformation. The stress release deformation δ1 and the cutting force elastic deformation δ2 are vector superimposed for pre-compensation, and the double closed loop is combined with intermediate secondary scanning actual measurement correction to control the influence of post-processing delay deformation within the range that can be absorbed by the finishing allowance (≤0.5mm), thereby reducing the number of rework and grinding.

[0046] (3) Ensure consistent surface quality on both sides of the interface. The use of partitioned force control parameter lookup table switching, dual-channel online identification of the interface, and pneumatic floating high-frequency buffer ensures that the normal force fluctuation after grinding in the transition zone of the additive / base material interface is ≤±5N, and the surface roughness on both sides meets Ra≤1.6μm, thus avoiding over-grinding or under-grinding.

[0047] (4) Eliminate the need for large component turnover and transportation. The in-situ processing mode eliminates the need for transportation, hoisting and secondary clamping of large frame structures, and significantly shortens the auxiliary time for a single piece; in the example, the typical time for the whole process is about 60 minutes, of which the clamping and alignment step is replaced by in-station scanning and registration.

[0048] (5) One set of equipment completes the entire process of measurement, milling and grinding. The three end faces of the quick-change disc share the same robot motion chain and coordinate system reference, avoiding reference transmission errors when switching between multiple devices. Attached Figure Description

[0049] Figure 1 This is a system architecture diagram of the additive forming connection area local in-situ flexible processing device of the present invention; in the figure, 1 is an omnidirectional moving platform, 101 is a Mecanum wheel omnidirectional moving mechanism, 102 is a hydraulic outrigger, 2 is an industrial robot, 201 is a robot coordinate system, 3 is an end effector group, 301 is a quick change plate, 302 is a 3D scanner, 303 is a milling spindle, 304 is an end flange, 4 is a dual compliant force control grinding head, 401 is a six-dimensional force sensor, 402 is a pneumatic floating mechanism, 403 is an eccentric vibrating grinding plate, 404 is an impedance controller, and 5 is a control system; Figure 2 This is a flowchart of the two-level registration and dual-channel residual distribution matrix generation method of the present invention; Figure 3 This is a flowchart of the stress relief and stiffness deformation dual pre-compensation and intermediate secondary scanning correction method of the present invention; Figure 4 This is a schematic diagram of the dual compliant force control grinding head of the present invention; Figure 5 This is a complete process flow diagram of the device of the present invention. Detailed Implementation

[0050] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings. It should be noted that the following embodiments are used to explain this invention, but do not constitute a limitation thereof.

[0051] The industrial robot of this invention is a KUKA KR500 R2830 MT, which conforms to ISO 9283 / GB / T 12642-2013.

[0052] The objects and boundary conditions of this invention are as follows: This embodiment focuses on the additive forming connection area of ​​an aircraft frame structure. Both the base material and the additive material are 5-series aluminum alloys, specifically grades 5052, 5083, and 5A06. Machining features include milling of the mating surfaces, milling of the mounting slots, contour trimming, and finishing of the stepped surfaces. The accuracy requirement is IT7~IT8 grade, corresponding to a tolerance of ±0.03~0.05mm. The surface roughness Ra after milling should not exceed 3.2μm, and after grinding, Ra should not exceed 1.6μm. The allowance deviation in the additive forming connection area is ±1.5~5mm, exhibiting an uneven distribution. Non-rigid deformation is localized warping of 0.5~2mm. The typical machining time for a single additive forming connection area is approximately 60 minutes.

[0053] Figure 1 This is a system architecture diagram of the additive forming connection area local in-situ flexible processing device of the present invention. It mainly includes: (1) Omnidirectional mobile platform 1: (2) Industrial robot 2: mounted on omnidirectional mobile platform 1.

[0054] (3) End effector group 3: Automatically switches between the following three end effects via pneumatic quick-change disc 301: 3D scanner 302, milling spindle 303, dual compliant force control grinding head 4.

[0055] (4) Dual compliant force control grinding head 4: It is composed of an end flange, a six-dimensional force sensor 401, a pneumatic floating mechanism 402 and an eccentric vibrating grinding disc 403 connected in series along the axial direction.

[0056] (5) Control system 5: includes a two-level registration and margin matrix module, an adaptive toolpath generation module, a stress release + stiffness dual pre-compensation module, a zoned force control and safety monitoring module, and a digital twin simulation verification module.

[0057] The key device parameters used in this embodiment are configured as follows: The industrial robot is a KUKA KR500 R2830MT with a rated load of 500kg, a maximum working radius of 2826mm, and a repeatability of ±0.08mm (compliant with ISO 9283 / GB / T12642-2013 standards). The omnidirectional moving platform 1 adopts a Mecanum wheel omnidirectional moving mechanism 101, which is operated remotely by hand and has four hydraulic outriggers 102. During processing, it lifts and unloads the load from the traveling wheels, with a load-bearing capacity of not less than 3.5 tons. The milling spindle 303 has a rated power of not more than 15kW (S1 continuous duty system), a maximum speed of not less than 24000rpm, an HSK-E50 tool interface, and cooling methods including water cooling, oil cooling, and MQL micro-lubrication. The 3D scanner 302 is preferably handheld and clamped at the end of the industrial robot, with a measurement accuracy of not more than 0.02~0.03mm. The pneumatic quick-change disc 301 has a repeatability of not more than ±0.02mm. The six-dimensional force sensor 401 has a sampling frequency of no less than 1000Hz. The pneumatic floating mechanism 402 has a floating stroke of ±5~8mm and a preload pressure adjustable within the range of 0.2~0.6MPa. The eccentric vibrating grinding disc 403 adopts an eccentric vibration type (Random Orbital Sander, ROS). The upper limit of the normal force is set at 150N; exceeding this limit triggers an emergency stop and drives the industrial robot to retreat along the normal direction. The reference ball used for hand-eye calibration and verification has a diameter of 25mm and a sphericity not exceeding 0.002mm, and is fixed to the omnidirectional moving platform surface.

[0058] The cutting parameters and zone force control parameters of this invention are as follows: This embodiment uses a Φ20mm end mill to mill 5-series aluminum alloy. The rough milling operation has a depth of cut ap of 2.0mm, a width of cut ae of 10mm, a feed rate F of 2300mm / min, a cutting speed Vc of 450m / min, a main cutting force of approximately 1840N, and a material removal rate of approximately 46cm³. 3 / min. The cutting depth ap of the finish milling process shall not exceed 0.5mm, the cutting width ae shall be 6~8mm, the feed rate F shall not exceed 1000mm / min, the cutting speed Vc shall be 500m / min, and the main cutting force shall not exceed 500N.

[0059] The dual-compliant force control grinding head switches force control parameters according to material zones, with a uniform virtual inertia M of 3 kg. The normal force in the additive manufacturing zone is set to 60~80 N, and the virtual stiffness K... s The aerodynamic preload pressure is 0.3 MPa, with a stiffness of 2000 N / m and a damping B of 300 N·s / m (low stiffness, high damping strategy). The normal force in the transition zone is set to 50~65 N. sWith a strength of 3000 N / m and a radius of 250 N·s / m, a stepped transition and feedforward buffering strategy is adopted, and the aerodynamic preload pressure is 0.4 MPa. The normal force in the base material region is set to 45~60 N, and K... s The pressure is 5000 N / m, B is 200 N·s / m (conventional impedance strategy), and the pneumatic preload pressure is 0.5 MPa. The above values ​​need to be calibrated according to the material removal rate test of the additive material and the base material.

[0060] Figure 5 This is a complete process flow diagram of the device of the present invention. The processing steps are as follows: the omnidirectional moving platform is driven to the workpiece station by manual remote control → the four hydraulic outriggers are lifted to unload the load of the traveling wheels → the quick-change plate switches the 3D scanner and completes the online verification of hand-eye calibration (online verification of hand-eye calibration: coincidence of the center of the reference ball) → automatic scanning of the additive forming connection area → two-level registration to generate a dual-channel allowance distribution matrix → adaptive toolpath generation and collision detection and digital twin simulation verification → the quick-change plate switches the milling spindle and performs dual pre-compensation rough milling (stress release δ1 and stiffness δ2) → intermediate secondary scanning and correction of the finishing toolpath → finish milling (IT7~IT8 grade, ±0.03~0.05mm) → the quick-change plate switches the dual compliant force control grinding heads and performs zoned force control grinding (Ra≤1.6μm) → final inspection scanning and qualification determination.

[0061] To address the issue of non-rigid deformation in the bonding area of ​​additive manufacturing, the implementation steps are as follows: Step 1: Manually remotely control the omnidirectional moving platform to the aircraft frame structure parking position, control the four hydraulic outriggers to lift the omnidirectional moving platform, unload the load on the traveling wheels, and complete the locking of the omnidirectional moving platform.

[0062] Step 2: The end effector of the industrial robot switches the 3D scanner via a pneumatic quick-change disc and performs online hand-eye calibration verification: The industrial robot holds the 3D scanner and automatically scans the connection area of ​​the miniature reference sphere on the platform surface in 3 to 5 postures. If the standard deviation of the sphere center does not exceed 0.05mm, it is considered qualified; otherwise, the AX=XB calibration is re-executed (N≥15 postures, solved by Tsai-Lenz method).

[0063] Step 3: The industrial robot automatically scans the additive forming connection area according to the planned trajectory to obtain the measured point cloud, and performs point cloud preprocessing, including noise reduction, simplification, and meshing, to obtain the meshed measured point cloud.

[0064] Step 4: Perform constrained weighted rigid body ICP registration and RBF / TPS non-rigid registration sequentially on the meshed measured point cloud; The first step involves using the tooling positioning surface and the machined surface of the base material as reference features. These features are assigned high matching weights. Noise denoising employs statistical outlier filtering (standard deviation factor set to 2.0). Simplification utilizes voxel grid downsampling (grid size 0.2~0.5mm, 0.3mm in this embodiment). Meshing involves reconstructing a triangular mesh surface from the simplified point cloud. Then, constrained weighted rigid body ICP registration is performed using the tooling positioning surface and the machined surface of the base material as reference features to solve for the six-degree-of-freedom rigid body transformation (R,t), where R is a 3×3 rotation matrix and t is a three-dimensional translation vector. Corresponding point pairs are established using nearest neighbor search, and outliers with a distance greater than 2 mm are removed. The matching weight of the baseline feature point set is 5 to 10 times that of the additive forming connection area point set (8 times in this embodiment) to prevent the undeformed area from being "averaged out" by the local warping of the additive forming connection area. Singular value decomposition (SVD) is used to solve the rigid body transformation (R,t) with a convergence condition that the change in rotation angle between two adjacent iterations is less than 10°. -5 rad and the translation change is less than 10 -4 mm; then fit the TPS deformation field to the additive forming connection area, and optimize the objective as the weighted sum of registration residual and bending energy.

[0065] The second paragraph discusses non-rigid registration. Radial Basis Functions (RBFs) are a class of basis functions whose values ​​depend only on the distance from a spatial point to a central control point; that is, their function values ​​are determined solely by |xc| and are independent of direction. A weighted superposition of RBFs at several control points can fit arbitrarily complex surfaces or their deformation fields. Thin Plate Splines (TPSs) are a specific form of RBFs, with a kernel function φ(r) = r. 2 ·log(r), geometrically equivalent to the shape with minimum bending energy after an infinitely thin metal plate is flexed at specified heights through several control points, is used. Therefore, TPS selects the smoothest surface among all possible surfaces passing through all control points. Its deformation function is... ,in This is the radial basis function of the TPS, where the independent variable r = ||xc|| represents the Euclidean distance from a spatial point x to a control point c, and c... i Let w be the i-th control point, i = 1, 2, ..., n, where n is the total number of control points. i The weights are to be determined and then processed by regularization. The bending energy is controlled, where λ is the regularization weighting coefficient used to balance the relative weight of the registration residual term and the bending energy term. A non-rigid deformation field refers to a displacement vector field in which relative displacement occurs between points within an object, changing its shape. This is distinct from a six-degree-of-freedom rigid body transformation that only involves translation and rotation. Local warping in the additive forming connection zone due to thermal residual stress is an example of this type of non-rigid deformation. In this embodiment, points in the additive forming connection zone with a registration residual greater than 0.1 mm after rigid body registration are used as TPS control points. The deformation field coefficients are solved by solving a closed-loop system of linear equations combining the residual and regularization terms. An anchoring constraint of f(x) = 0 is applied to all reference feature points, ensuring that f(x) only absorbs the local warping deformation in the additive forming connection zone without affecting the global reference already locked by rigid body registration.

[0066] Step 5: Generate a dual-channel margin distribution matrix: The first channel represents the geometric margin h(x) (the signed distance along the tool axis from the theoretical profile to the measured surface), and the second channel represents the material partitioning identifiers for the additive manufacturing zone, transition zone, and base material zone. The material partitioning map is generated using the following method: the additive manufacturing zone outline is taken from the deposition boundary given by the additive deposition path planning data; the transition zone is a ring formed by extending the additive manufacturing zone outline outward along the surface normal by 5-10 mm (8 mm in this embodiment); the remaining areas are identified as the base material zone. After the two-level registration is completed, the material partitioning map is superimposed on the measured point cloud in the same coordinate system, and the partition to which it belongs is queried for each tool position.

[0067] Step 6: Based on the dual-channel margin distribution matrix, select cutting parameters by looking up a table according to the two dimensions of "margin value × material zone", and automatically generate an adaptive toolpath; then perform collision detection and digital twin simulation verification. Figure 2 This is a flowchart of the two-level registration and dual-channel margin distribution matrix generation method of the present invention. The robot's posture is confirmed to be collision-free and the end effector's overall deformation is within acceptable limits. The digital twin model used for the digital twin simulation verification includes a robot kinematics and dynamics model, a measured point cloud model of the additive forming connection area after two-level registration, and an empirical cutting force model. The toolpath execution process is reproduced in the simulation environment, and the end effector pose deviation and collision check results for each tool position point are output.

[0068] Step 7: Switch the milling spindle and perform double pre-compensation rough milling: finite element layer removal simulation predicts stress release deformation δ1, the whole machine stiffness model estimates cutting force elastic deformation δ2, and the tool position is set according to... The reverse offset compensation is calculated by taking the component of the overall deformation in the normal direction of the machined surface.

[0069] Step 8: After rough milling is completed, switch back to the 3D scanner, scan the rough-milled surface, and measure the residual deviation after stress release; if the residual deviation is greater than 0.05mm, correct the finishing toolpath; otherwise, directly perform finish milling.

[0070] Step 9: Perform finish milling with a cutting depth of ap not exceeding 0.5mm to achieve IT7~IT8 level accuracy. Figure 3 This is a flowchart of the dual pre-compensation and intermediate secondary scanning correction method of the present invention.

[0071] Step 10: Switch to dual compliant force control polishing heads. Figure 4 This is a schematic diagram of the dual compliant force-controlled grinding head of the present invention. Grinding is performed according to the cutting parameters and zoned force control parameters of the above-described embodiment of the present invention. During the grinding process, the additive / base material interface is identified online through a dual-channel method using the prior position of the material zoning diagram and the force signal spectrum characteristics, and the force control parameter group is switched in advance. Specifically, the dual compliant force-controlled grinding head is connected to the end flange of the robot; it specifically includes a six-dimensional force sensor located downstream of the end flange, the sampling frequency of the six-dimensional force sensor being not less than 1000Hz; a pneumatic floating mechanism located downstream of the six-dimensional force sensor, the pneumatic floating mechanism having a floating stroke of ±5~8mm, and the cylinder preload pressure adjustable within the range of 0.2~0.6MPa; an eccentric vibrating grinding disc located downstream of the pneumatic floating mechanism; and an impedance controller used to control the deviation between the normal contact force measured by the six-dimensional force sensor and the target normal force in the zoned force control parameter table, according to a control cycle of 1kHz and based on a second-order impedance model. Solve for the position correction Δx and superimpose it onto the original grinding trajectory, where F n,des For the target normal contact force in the force control parameter table of the partition, F n,meas The normal contact force is measured by the six-dimensional force sensor; the partitioned force control parameter table stores the normal force setting value, virtual inertia M, damping B, and virtual stiffness K in the additive manufacturing zone, transition zone, and parent material zone, respectively. s The pneumatic preload pressure and the position correction of the impedance controller are ultimately superimposed on the movement of the pneumatic floating mechanism, which is the final displacement of the dual compliant force control grinding head. The interface recognition module is used to identify the additive / base material interface online during grinding based on the prior interface position and / or the spectral characteristics of the normal force signal given by the scanned material partition map. The quantitative criterion for spectral recognition is: when the abrupt change in the spectral energy of the normal force signal in the 50~200Hz frequency band relative to the previous steady-state measurement segment (time window 0.5~1s) exceeds 20%, it is determined that the grinding wheel has crossed the interface, and upon interface recognition, the impedance controller and pneumatic floating mechanism are triggered to switch to the force control parameters of the corresponding partition. Safety monitoring triggers an emergency stop and drives the industrial robot to retreat along the normal direction when the normal contact force exceeds the hard upper limit of 150N. The material partition map is scanned to verify the interface position.

[0072] Step 11: Switch the 3D scanner again using the pneumatic quick-change disc, perform a final inspection scan on the processing area, output a deviation chromatogram, and determine the pass / fail status.

[0073] Example 2

[0074] When the additive forming connection area contains only the weld repair area and has no obvious material partition gradient, the material partition channel degenerates into a single partition, while the other method steps (two-level registration, double pre-compensation, and intermediate secondary scanning) remain unchanged, reflecting the compatibility and scalability of the solution.

[0075] Example 3

[0076] By adjusting the cutting tools and cutting parameters to process parameters suitable for titanium alloys (e.g., reducing the cutting speed Vc to 60~120m / min), the method chain described in this invention can be transferred to the processing of the titanium alloy additive forming connection area, while the device composition remains unchanged.

[0077] Example 4

[0078] When there are multiple additive forming connection areas for the same component, the platform will be transferred to each workstation in sequence to execute steps 2 to 11 independently; hand-eye calibration verification is only enforced on the first start of each day or after the 3D scanner is replaced.

[0079] Comparative Example The existing method for handling the additive forming connection area—which involves directly scanning the entire connection area and generating a program based on a single allowance threshold for continuous milling and grinding—is flawed because it lacks pre-defined material zoning force control. The grinding process applies the same normal force setting to the additive area, transition area, and base material area, failing to adapt to material properties. The additive area has lower hardness and a stronger tendency for work hardening, resulting in a higher material removal rate under the same normal force, easily leading to overcutting and surface pits. Conversely, the base material area has a denser structure and higher hardness, resulting in a lower material removal rate under the same normal force, making it difficult to meet surface quality requirements. Furthermore, this method immediately solidifies the toolpath after scanning and executes continuously without secondary scanning or toolpath correction. The delayed deformation caused by residual stress release during milling is not reflected in the one-time generated program and can only be detected after grinding. At this point, no subsequent processes can compensate, requiring rescanning, regenerating the toolpath, and remapping, thus compromising the already achieved dimensional accuracy and increasing process cycle time and tool consumption.

Claims

1. A method for local in-situ flexible processing of the bonding area in additive manufacturing, characterized in that, Includes the following steps: S1: Scan The measured point cloud was obtained by scanning the additive forming connection area using a 3D scanner held at the end of an industrial robot. The measured point cloud was then preprocessed by denoising, simplification, and meshing to obtain a meshed measured point cloud. S2: Two-level registration Using the tooling positioning surface and the machined surface of the base material as reference features, constrained weighted rigid body iterative nearest point (ICP) registration is performed between the meshed measured point cloud and the theoretical CAD model. Corresponding point pairs between the measured points and the theoretical model are established through nearest neighbor search, and outliers with a distance greater than 2 mm are removed. The weighted registration objective function is constructed using the sum of the squares of the distances between corresponding point pairs. The matching weight of each point in the reference feature point set is taken as 5 to 10 times the matching weight of each point in the additive forming connection area point set. Singular value decomposition (SVD) is used to find the six-degree-of-freedom rigid body transformation that minimizes the weighted registration objective function, with the rotation angle change between two adjacent iterations being less than 10°. -5 rad and the translation change is less than 10 -4 mm is used as the convergence condition for the iteration; Based on the rigid body registration results, a non-rigid deformation field of a thin plate spline (TPS) based on radial basis function (RBF) is superimposed on the point cloud of the additive forming connection area. With the minimum residual registration error and the minimum bending energy as the optimization objectives, the deformation field coefficients are solved to obtain a two-level registration model that includes rigid body transformation and non-rigid deformation field. S3: Dual-channel margin distribution matrix The measured point cloud is aligned to the theoretical digital model based on the two-level registration model, and a dual-channel margin distribution matrix is ​​generated: the first channel is the geometric margin value h(x) of each scanned spatial point, where x is the three-dimensional position vector of the scanned spatial point; the second channel is the material partition identifier at the corresponding tool point, which includes the additive region, transition region and parent material region. S4: Adaptive Toolpath Based on the dual-channel margin distribution matrix, the cutting parameters are selected by looking up a table in the two dimensions of "margin value × material zone", and an adaptive toolpath for milling the connecting area of ​​additive forming is automatically generated.

2. The method for local in-situ flexible processing of the additive forming connection area according to claim 1, characterized in that, The reference features for the constraint-weighted rigid body iterative nearest point (ICP) registration include at least one of the following: tooling positioning surface, machined surface of the base material, fixture reference pin hole, and platform table reference point calibrated by a coordinate measuring machine.

3. The method for local in-situ flexible processing of the additive forming connection area according to claim 1, characterized in that, The non-rigid deformation field adopts a thin plate spline (TPS) model, and its deformation function is: ,in, This is the radial basis function of the TPS, where the independent variable r = ||xc|| represents the Euclidean distance from a spatial point x to a control point c, and c... i Let w be the i-th control point, i = 1, 2, ..., n, where n is the total number of control points. i The weights are to be determined and then processed by regularization. The bending energy is controlled, where λ is the regularization weighting coefficient and x is the three-dimensional position vector of the spatial point obtained by scanning, which is used to balance the relative weight of the registration residual term and the bending energy term.

4. The method for local in-situ flexible processing of the additive forming connection area according to claim 1, characterized in that, The rule for selecting cutting parameters based on a two-dimensional lookup table using "allowance value × material zone" is as follows: when the geometric allowance h > h 精 When the material is in the additive manufacturing zone or transition zone, rough machining is performed with a depth of cut ap = 2.0 mm, a width of cut ae = 10 mm, feed rate F = 2300 mm / min, and cutting speed Vc = 450 m / min; when 0 <h≤h 精 When performing finishing, ap ≤ 0.5mm, F ≤ 1000mm / min; when h ≤ 0, mark it as empty cut and skip the tool position; finishing allowance threshold h. 精 Take 0.5mm.

5. The method for local in-situ flexible processing of the additive forming connection area according to claim 1, characterized in that, It also includes: after the adaptive toolpath is generated, collision detection and digital twin simulation verification are performed on the robot to confirm that there is no collision and the predicted end-effector deformation does not exceed 0.1mm before milling is performed.

6. The method for local in-situ flexible processing of the additive forming connection area according to claim 1, characterized in that, It also includes a pre-compensation method for machining deformation in the additive forming connection zone, considering the coupling of residual stress relief and stiffness, comprising the following steps: S5: Double pre-compensation rough milling Based on the dual-channel residual distribution matrix obtained from the scan, a local finite element model of the additive forming connection area is established. The initial residual stress field in the finite element model is assigned values ​​according to the additive process simulation results or measured residual stress data. The removed elements are gradually deleted according to the layered removal sequence of the roughing toolpath, and the stress release deformation field δ1(x) caused by residual stress release in each processing stage is simulated and predicted. A static stiffness model of the whole machine, including the stiffness of the hydraulic outriggers of the mobile platform, the stiffness of the robot joints, and the stiffness of the linkage structure, is established. The robot joint angle vector q corresponding to the current tool position point and the end cutting force vector F are then compared. c Substitute the Cartesian stiffness matrix K(q) at the end of the machine to estimate the elastic deformation field δ2(x) induced by the cutting force; transform the stress release deformation field δ1(x) and the elastic deformation field δ2(x) of the cutting force to the robot base coordinate system and then perform vector superposition. Take the component of the superposition result in the normal direction of the machining surface as the comprehensive deformation prediction value, and shift the tool position coordinates in the opposite direction of the component in the toolpath post-processing stage to generate a rough milling toolpath with double pre-compensation; S6: Intermediate Secondary Scan Correction After rough milling and before finish milling, the surface is left to stand until the surface deformation increment measured twice is less than the set threshold. Then, an intermediate second scan is performed to rescan the rough-milled surface and measure the actual residual deviation after stress release. When the measured residual deviation is greater than the set threshold, the finish machining toolpath is corrected according to the measured deviation. After completing the intermediate second scan and finish machining toolpath correction, the milling spindle is switched back through the pneumatic quick-change disc to perform finish milling with a cutting depth of ap≤0.5 mm. When the residual deviation is greater than the set threshold, the finish machining toolpath is corrected according to the measured deviation, forming a "prediction + measurement correction" double closed loop.

7. The method for local in-situ flexible processing of the additive forming connection area according to claim 6, characterized in that, The Cartesian stiffness matrix K(q) of the end effector in the overall static stiffness model varies with the robot joint angle vector q, and is composed of the robot Jacobian matrix J(q) and the joint stiffness diagonal matrix K. θ According to the mapping relationship K(q)=J - ᵀ·K θ ·J -1 The calculation yields J, where J T It is the transpose of the Jacobian matrix, K θ J is the joint stiffness diagonal matrix. -1 It is the inverse of the Jacobian matrix; and it is synthesized with the stiffness of the hydraulic outrigger and the stiffness of the linkage structure according to the series and parallel spring model; the stiffness of the linkage structure is obtained through finite element analysis, and the stiffness of the robot joint is obtained through technical specifications or stiffness identification experiments; the cutting force vector is obtained based on the empirical cutting force model or finite element cutting simulation established by tool diameter, cutting depth, cutting width, feed rate and cutting speed.

8. The method for local in-situ flexible processing of the additive forming connection area according to claim 1, characterized in that, Also includes S7: Quick-change disc switching to dual-compliance force control polishing heads and performing zoned force control polishing. The dual compliant force control grinding head adopts an architecture that couples active impedance control with passive pneumatic floating, and introduces a material partition force control parameter lookup and switching mechanism. Active impedance control includes a six-dimensional force sensor and an impedance controller. The impedance controller is used to determine the deviation between the normal contact force measured by the six-dimensional force sensor and the target normal force in the partitioned force control parameter table. The six-dimensional force sensor measures the normal contact force in real time, and the impedance controller operates according to a second-order impedance model with a control cycle of 1 kHz. The normal force deviation is mapped to a position correction Δx and superimposed on the original grinding trajectory, where M is the virtual inertia, B is the damping, and K is the position correction value. s For virtual stiffness, F n,des For the target normal force in the force control parameter table of the partition, F n,meas The normal contact force was measured using a six-dimensional force sensor. The second derivative of the position correction is... The first derivative of the position correction is... This is the position correction amount, and this value is constantly changing; The normal force setting value and impedance parameter are switched according to the preset material partition force control parameter table. The partition force control parameter table stores the normal force setting value, virtual inertia M, damping B, and virtual stiffness K for the additive manufacturing zone, transition zone, and base material zone, respectively. s And pneumatic preload pressure; The passive pneumatic floating mechanism is a pneumatic floating mechanism connected in series between the six-dimensional force sensor and the eccentric vibrating grinding disc. The floating stroke is ±5~8mm, and the cylinder preload pressure is switched synchronously with the material partition within the range of 0.2~0.6MPa. Active impedance control focuses on low-frequency large deviation compensation, while passive aerodynamic floating absorbs high-frequency force disturbances outside the active impedance control bandwidth. In the event of active impedance control failure, passive aerodynamic floating can still maintain acceptable contact force, thus providing safety redundancy. S8: Final inspection scan and pass / fail determination The final inspection scan is an interface recognition method, employing a dual-channel redundancy mechanism: Channel 1 uses the prior interface position and / or normal force signal spectrum characteristics given by the scanning material partition map to identify the additive / base material interface online, triggering parameter switching along the grinding trajectory position; Channel 2 detects the spectrum characteristics of the normal force signal in the 50~200Hz frequency band online, with the spectrum energy calculated using a 0.5~1s sliding time window. When the abrupt change exceeds 20%, it is determined that the grinding wheel has crossed the interface. When the grinding wheel crosses the interface, the force fluctuation spectrum characteristics change abruptly. After real-time discrimination, it is cross-verified with the prior position, triggering the impedance controller and pneumatic floating mechanism to switch to the force control parameters of the corresponding partition.

9. The method for local in-situ flexible processing of the additive forming connection area according to claim 8, characterized in that, The normal force setting value is 60~80N in the additive manufacturing zone, 50~65N in the transition zone, and 45~60N in the base material zone; virtual stiffness K s The transition zone is lower than the base material zone to buffer the sudden force change when the grinding wheel crosses the interface.

10. A local in-situ flexible processing device for the bonding area of ​​additive manufacturing, characterized in that, The additive manufacturing connection area local in-situ flexible machining device includes an omnidirectional moving platform, an industrial robot, an end effector assembly, dual compliant force-controlled grinding heads, and a control system. The industrial robot is mounted on the omnidirectional moving platform to move to the workpiece parking position. The end effector assembly is connected to the end effector assembly. The end effector assembly automatically switches between three end effectsors via a quick-change disc: a 3D scanner, a milling spindle, and dual compliant force-controlled grinding heads. The dual compliant force-controlled grinding heads consist of a six-dimensional force sensor, a pneumatic floating mechanism, and an eccentric vibrating grinding disc connected in series along the axial direction. The control system is used to automate the local in-situ flexible machining of the additive manufacturing connection area, including the local in-situ flexible machining method, the pre-compensation method for machining deformation, and the switching and movement of the three end effectors.

11. The additive forming connection area local in-situ flexible processing device according to claim 10, characterized in that, The omnidirectional mobile platform adopts a Mecanum wheel omnidirectional mobile mechanism, is equipped with a manual remote control operating system, and has four hydraulic outriggers. During processing, the four hydraulic outriggers are raised to unload the load from the traveling wheels; the load-bearing capacity is not less than 3.5 tons. And / or, the industrial robot has an end flange and a robot base coordinate system, a rated load of 500 kg, a maximum working radius of 2826 mm, and a repeatability of ±0.08 mm; And / or, the quick-change disc is a pneumatic quick-change disc with a repeatability accuracy ≤ ±0.02mm, a 3D scanner measurement accuracy ≤ 0.02~0.03mm, a milling spindle rated power ≤ 15kW, S1 continuous duty cycle, maximum speed ≥ 24000rpm, HSK-E50 tool interface, water cooling + MQL; And / or, the sampling frequency of the six-dimensional force sensor is not less than 1000Hz; the pneumatic floating mechanism has a floating stroke of ±5~8mm, and the cylinder preload pressure is adjustable in the range of 0.2~0.6MPa; And / or, the control system includes a two-level registration and allowance matrix module, an adaptive toolpath generation module, a stress release + stiffness dual pre-compensation module, a zoned force control and safety monitoring module, a digital twin simulation verification module, and a hand-eye calibration verification module; the two-level registration and allowance matrix module is used to obtain and provide feedback on the machining allowance; the adaptive toolpath generation module is used to determine and provide feedback on the robot machining trajectory; the stress release + stiffness dual pre-compensation module is used to provide part accuracy compensation; the zoned force control and safety monitoring module is used to provide safe and compliant grinding; the digital twin simulation verification module is used to provide virtual simulation of the entire machining process; the hand-eye calibration verification module is used to scan and check the overlap of the ball center position using a miniature reference ball fixed to the moving platform surface in 3-5 postures after each start-up or replacement of the 3D scanner, and prompts recalibration when the standard deviation of the ball center exceeds 0.05mm.

12. The additive forming connection area local in-situ flexible processing device according to claim 10, characterized in that, The pneumatic floating mechanism includes a cylinder, a linear guide, and a preload spring installed in series. The cylinder preload pressure is adjusted in real time by a proportional pressure regulating valve according to the corresponding value in the partition force control parameter table.

13. The additive forming connection area local in-situ flexible processing device according to claim 10, characterized in that, It also includes a safety monitoring module, which triggers an emergency stop and drives the industrial robot to retreat along the normal direction when the normal contact force exceeds the hard upper limit of 150N.

Images