A five-axis machining method for a robot thin-walled structure

Abstract

The present application relates to the technical field of robot high-precision machining and multi-axis machining of complex thin-walled structural parts, and particularly relates to a robot thin-walled structural part five-axis machining method, which comprises determining the quasi-static flexibility and dynamic response of the thin-walled region respectively according to the actual material removal state, and predicting the subsequent stiffness drop region to complete the robot posture migration in advance; for the candidate tool axis and the robot redundant posture, the cutting balance state is obtained by iteration of the cutting contact region, the cutting load and the tool-workpiece relative deformation, and the stable machining state and the allowable feed interval are determined in combination with the three-dimensional dynamic response; after the tool position and the tool axis are jointly compensated, the cutting contact state, the deformation and the stability are rechecked. The present application synchronously collects the cutting force, the vibration and the machining result in the machining, selectively updates the model according to the deviation source, and reconstructs the subsequent machining state when the material state changes or the associated state deviates together, thereby improving the robot five-axis machining precision and stability of the thin-walled structural part.

Description

[Technical Field] This invention relates to the field of bonding equipment roller technology, and in particular to a five-axis machining method for thin-walled structural parts for robots. [Background Technology] Thin-walled structural components are characterized by their light weight, high specific strength, and high structural integration, and are widely used in aerospace, high-end equipment, and transportation equipment. These components are typically formed from blanks with a large machining allowance through multi-stage material removal. As material is continuously removed, the remaining wall thickness, local boundary conditions, mass distribution, and stiffness of the part all change, causing significant time-varying characteristics in the workpiece's deformation and vibration during processing.

[0001] Traditional five-axis CNC machine tools have high structural rigidity and motion accuracy, but large five-axis machining equipment is expensive, and the workspace is limited by the machine tool structure and stroke. Industrial robots have the advantages of large workspace, high flexibility, and easy reconfiguration. After installing a high-speed machining spindle at the end effector, they can complete multi-axis machining of complex curved surfaces, and are especially suitable for local machining of large components and multi-variety machining scenarios.

[0002] However, the joint transmission mechanisms and linkage systems of serial industrial robots exhibit significant flexibility, and their end effector stiffness typically changes with the robot's joint posture. For a five-axis machining task performed by a six-degree-of-freedom robot, after determining the tool center position and tool axis direction, the robot can also change the position of each joint by altering redundant postures around the tool axis direction. Therefore, the same tool position and tool axis direction can correspond to different robot machining stiffness and dynamic response.

[0003] On the other hand, thin-walled structural components inherently have low stiffness. When a robot processes a thin-walled structural component, the cutting load simultaneously causes tool-side deformation (consisting of the robot, machining spindle, and cutting tool) and thin-walled structural component-side deformation. The final machining profile is not determined solely by the absolute deformation on either side, but rather by the relative deformation between the cutting tool and the workpiece.

[0004] Existing technologies have addressed these issues from various perspectives.

[0005] For example, Chinese invention patent application CN115229796A discloses a robot end-effector pose tracking and compensation method for milling large aerospace components. The robot stiffness at the milling contact position is determined by identifying robot joint stiffness parameters and converting flexibility characteristics. The processing error caused by robot flexibility is reduced by redundancy angle optimization, cutting force prediction, pre-compensation and real-time pose compensation.

[0006] Chinese invention patent application CN114833848A discloses a rigidity-constrained robot milling tool axis vector and redundancy integrated planning method, which improves the robot milling posture by jointly planning the tool axis vector and robot redundancy angle.

[0007] Chinese invention patent application CN117908466A discloses a method for optimizing the tool axis direction and redundancy angle in robot ball end milling. After changing the tool axis direction, the tool-workpiece contact area and cutting force are redefined, and the tool tilt angle, side tilt angle and robot redundancy angle are jointly optimized.

[0008] Chinese invention patent application CN107942936A discloses a method for compensating for the deformation error between a five-axis side milling tool and a workpiece. The method calculates the tool deformation and workpiece deformation, and performs iterative calculations considering the feedback effect of deformation on the milling force to compensate for the tool tip position and tool axis tilt angle.

[0009] Chinese invention patent application CN111723440A discloses a hybrid modeling method for predicting the machining accuracy of thin-walled parts. It obtains the time-varying compliance during the machining process through dynamic material removal and combines the workpiece end deformation and tool end deformation to predict the machining accuracy of thin-walled parts.

[0010] The above technologies have solved problems such as robot posture optimization, robot end-effector deformation, tool and workpiece deformation compensation, and time-varying compliance prediction of thin-walled parts, but the following technical problems still exist in the five-axis machining of thin-walled structural parts by robots.

[0011] First, changing the tool axis posture alters the actual contact area between the tool and the workpiece, the position of the cutting edge involved in the cutting, and the cutting load. Furthermore, after the robot and thin-walled structural components deform under the cutting load, the actual tool-workpiece contact area changes again. If the cutting load is calculated only once based on rigid geometry and the deformation is directly predicted, a consistent machining equilibrium cannot be achieved between the cutting load, the contact area, and the deformation.

[0012] Secondly, after modifying the tool position or tool axis direction based on the predicted deformation, the compensated tool-workpiece contact area will change again. If the machining is performed directly using the cutting load and stability results before compensation, the compensation result may be inconsistent with the re-formed actual machining state.

[0013] Furthermore, the quasi-static elastic deformation of thin-walled structural components and the dynamic vibration during processing are different physical responses. The former mainly affects the contour error, while the latter mainly affects processing stability and the actual dynamic contact state. If a single "stiffness index" is used to describe both responses simultaneously, it can easily lead to ambiguity in the physical meaning between error compensation and dynamic stability evaluation.

[0014] Furthermore, the causes of deviations in actual cutting loads, vibrations, and machining profiles are not the same. Cutting load deviations may originate from changes in tool condition or cutting parameters; vibration changes may stem from variations in system dynamic parameters; and deviations in actual wall thickness or material boundaries directly alter the subsequent workpiece structure. If all models are modified simultaneously upon the occurrence of any measurement deviation, not only will the computational load be substantial, but it is also prone to propagating the error to unrelated models.

[0015] In the later stages of the processing, when continuous material removal causes a rapid decrease in stiffness in thin-walled regions, if the robot's posture is only changed after entering the low-stiffness region, the posture transition process itself may cause abrupt changes in joint movement and processing state. Therefore, it is necessary to predict the subsequent structural state before entering the region where stiffness decreases significantly, and to complete the robot's posture transition in advance within the relatively high-stiffness processing region.

[0016] Therefore, there is a need for a five-axis machining method for thin-walled robotic structures that can distinguish between quasi-static deformation and dynamic response, achieve a consistent cutting equilibrium state in the tool-workpiece contact area, cutting load, and relative deformation, and can re-verify the compensated state, selectively update the model based on the source of deviation, and plan the robot's posture migration in advance based on future stiffness changes. [Summary of the Invention] To overcome the above problems, this invention proposes a five-axis machining method for thin-walled robotic structural parts that can effectively solve the above problems.

[0017] The present invention provides a technical solution to solve the above-mentioned technical problems: a five-axis machining method for thin-walled structural parts using an industrial robot with at least six degrees of freedom, carrying a machining spindle and milling cutters to perform five-axis linkage machining on the thin-walled structural parts, comprising the following steps: Step S1: Obtain the initial geometric model, target machining profile, blank state, clamping constraints and machining allowance of the thin-walled structural part to be machined, establish a material removal state model that characterizes the spatial distribution of the current remaining material, and divide the predetermined five-axis toolpath into multiple machining segments to be executed sequentially based on the curvature change of the surface to be machined, the change of the remaining wall thickness and the change of the theoretical tool axis direction. Step S2: Before executing the current processing segment, determine the remaining wall thickness, local boundary constraints, and remaining material distribution of the current processing segment based on the material removal state model, and establish a local quasi-static compliance state for contour error prediction and a local dynamic response state for vibration stability judgment, respectively; at the same time, predict the local dynamic response state after material removal of the subsequent processing segments according to the predetermined processing sequence; when the predicted local stiffness drop between adjacent subsequent processing segments reaches the preset sudden drop condition, determine the stiffness drop region and the attitude advance migration interval located upstream of the stiffness drop region; Step S3: Under the constraints of no interference between the tool and the workpiece, the range of motion of the robot joints, the robot's singular configuration avoidance, and the attitude continuity, the tool's forward tilt angle, the tool's side tilt angle, and the redundant attitude of the robot around the tool axis are changed to form multiple candidate machining states. For each candidate machining state, the rigid geometric contact area between the tool and the thin-walled structure is re-determined according to the current material removal state, and the effective cutting edge position, material removal state, and cutting speed state are determined according to the rigid geometric contact area to obtain the corresponding initial predicted cutting load. Step S4: Determine the quasi-static displacement of the tool side based on the initial predicted cutting load, the compliance state corresponding to the robot's current candidate posture, and the compliance state of the machining spindle and milling tool. Determine the quasi-static displacement of the thin-walled structure side based on the local quasi-static compliance state of the current machining segment. Transform the quasi-static displacement of the tool side and the quasi-static displacement of the thin-walled structure side to the same workpiece coordinate system to determine the quasi-static relative deformation between the tool and the thin-walled structure. Correct the actual spatial relationship between the tool and the thin-walled structure based on the quasi-static relative deformation, redetermine the cutting contact area and cutting load, and recalculate the quasi-static relative deformation until the changes in cutting load and normal relative deformation in two adjacent calculations both meet the preset convergence conditions. Determine the converged cutting contact area, cutting load, and quasi-static relative deformation as the cutting equilibrium state of the corresponding candidate machining state. Step S5: Based on the cutting balance state, determine the dynamic responses of the tool side and the thin-walled structural part side along the normal direction of the machined surface, the feed direction, and the tool axis direction, respectively. Judge the machining stability based on the relative dynamic responses of the two. When the relative vibration in the tool axis direction reaches a level that can change the actual contact range of the tool axis, redetermine the dynamic cutting contact area affected by the vibration, and recalculate the dynamic cutting load and dynamic stability state based on the changed actual contact state. At the same time, determine the allowable feed range corresponding to each candidate machining state based on the robot joint motion capability and the allowable vibration state of the thin-walled structural part. Step S6: Eliminate candidate machining states that do not meet the contour error requirements, dynamic stability requirements, no interference requirements, robot motion constraints, or have no effective allowable feed range; screen the remaining candidate machining states in the order of machining accuracy, dynamic stability margin, allowable feed capacity, and attitude continuity of adjacent machining segments, and plan the machining state sequence for multiple consecutive machining segments; when the attitude early migration interval exists, allocate the redundant robot attitude and tool axis attitude change required after entering the stiffness drop region to the consecutive machining segments within the attitude early migration interval, so that the robot completes at least the main part of the attitude transformation before entering the stiffness drop region; Step S7: Based on the determined target machining state, compensate the theoretical five-axis toolpath. Specifically, determine the tool position compensation amount based on the normal relative deformation of multiple representative feature positions within the effective cutting area, determine the tool axis direction compensation amount based on the changing trend of the normal relative deformation between each representative feature position along the tool axis direction, and determine the target feed state within the allowable feed range to form a compensated toolpath to be verified that includes tool position, tool axis direction, and time-varying feed information. Step S8: For the toolpath to be checked and compensated, redetermine the compensated tool-workpiece contact area, cutting load, quasi-static relative deformation, and dynamic stability. If the residual contour error, dynamic stability, tool-workpiece non-interference state, and robot motion continuity do not simultaneously meet the requirements, correct the tool position, tool axis direction, or machining parameters again, and recheck. When all the check conditions are met simultaneously, determine the corresponding toolpath as the five-axis toolpath for the current machining segment. Step S9: Control the robot to complete the current machining segment according to the five-axis toolpath. During the machining process, simultaneously collect the actual cutting load, robot motion state, and vibration response that correspond to the current machining segment and tool position. After completing the predetermined machining segment, obtain at least one of the following machining results: actual machining contour, actual wall thickness, or actual material removal boundary. Based on the corresponding deviation between the actual data and the predicted data, identify the source of the deviation and selectively correct the cutting load prediction state, local dynamic response state, or material removal state model according to the source of the deviation. When the material removal state is corrected, or multiple related states simultaneously show a continuous deviation that reaches the preset reconstruction condition, redetermine the local quasi-static compliance state, local dynamic response state, tool-workpiece contact area, candidate tool axis direction, robot redundant posture, allowable feed range, and toolpath compensation amount of the subsequent machining segment, starting from the corrected actual material removal state. Step S10: Perform subsequent machining segments according to the updated machining status until the five-axis machining of the thin-walled structural component is completed.

[0018] Preferably, when dividing the five-axis toolpath into machining segments in step S1: The tool position is obtained sequentially along the theoretical five-axis toolpath; The curvature of the machined surface, the theoretical remaining wall thickness, and the theoretical tool axis direction at each tool position are obtained respectively. When at least one of the surface curvature change, theoretical remaining wall thickness change, and theoretical tool axis direction change between adjacent tool positions meets the corresponding preset segmentation conditions, the machining segment boundary is established. The material removal state, local structural state, and tool-workpiece contact state within the same processing section are kept within a predetermined range of variation.

[0019] Preferably, in step S2: The local quasi-static compliance state includes at least the unit load displacement response of the current machining region along the normal to the machined surface and the direction of the main cutting load; The local dynamic response state includes at least the dynamic compliance characteristics of the current machining area along the normal of the machined surface, the feed direction and the tool axis direction, at least one low-order natural frequency and the corresponding damping characteristics. Multiple material removal state samples are pre-established based on different remaining wall thicknesses, local boundary constraints, and remaining material distributions, and the local quasi-static compliance state and local dynamic response state corresponding to each material removal state sample are obtained. During the formal processing, the corresponding state sample is selected based on the current material removal state, or the initial state parameters of the current processing area are obtained based on adjacent state samples, and the state parameters corresponding to the subsequent processing sections are corrected using actual processing data.

[0020] Preferably, determining the attitude advance migration interval in step S2 includes: Based on the current actual material removal status and the predetermined processing sequence, the predicted remaining material status after the completion of multiple future processing stages is generated sequentially. The predicted local dynamic response state of the corresponding region is determined based on the predicted remaining material state of each region. Compare the local stiffness changes between adjacent future processing sections; When the stiffness decrease or the stiffness decrease rate reaches the preset sudden drop condition, the corresponding processing section will be identified as the stiffness sudden drop region. Search upstream of the stiffness drop region along the processing direction for a processing segment that can meet the requirements of continuous robot posture change and processing stability, and determine the continuous processing segment obtained by the search as the posture advance migration interval.

[0021] Preferably, before formal processing, multiple calibration poses are selected in the robot workspace covering the actual processing area, and known loads in multiple directions are applied to the robot end effector and the pose changes of the robot end effector are measured to obtain the equivalent stiffness parameters of the robot joints. For each candidate machining state, the robot end effector flexibility is obtained based on the current joint position of the robot and the equivalent stiffness parameters of the robot joint, and the robot end effector flexibility is converted to the tool-workpiece contact area; The equivalent compliance of the machining spindle and milling cutter is added to the converted robot end effector compliance to determine the overall compliance state of the tool side. The tool side quasi-static displacement includes at least the translation of the robot end effector, the additional displacement of the tool position caused by the rotation of the robot end effector, the elastic deformation of the machining spindle, and the bending deformation of the milling tool.

[0022] Preferably, in step S3, redetermining the tool-workpiece contact area for each candidate tool axis direction includes: Determine the boundary of the remaining material entity based on the current material removal state model; Determine the position where the cutting edge enters the remaining material, the position where it leaves the remaining material, and the effective cutting range along the tool axis in the current candidate tool axis direction; For ball end mills, the actual spherical area involved in cutting is determined based on the positional relationship between the spherical cutting edge and the remaining material entity. The corresponding cutting speed state and instantaneous material removal state are determined based on the actual position of the cutting edge involved in the cutting. The cutting load is determined separately for different candidate tool axis directions, and the same cutting load is not directly used for deformation evaluation in different candidate tool axis directions.

[0023] Preferably, the cutting balance state in step S4 is determined in the following manner: The cutting load obtained based on the rigid geometric contact region is used as the first predicted cutting load; The quasi-static displacement of the tool side and the quasi-static displacement of the thin-walled structural component side are determined by the first predicted cutting load. The predicted actual position of the tool relative to the thin-walled structure is changed based on the quasi-static relative deformation of the two, and the tool-workpiece contact area and predicted cutting load for the next cutting are redefined. Repeatedly perform the update of the contact area, the update of the cutting load, and the update of the quasi-static relative deformation; The iteration ends when the representative cutting load change and normal relative deformation change obtained in two consecutive calculations are both less than the corresponding convergence allowable value. If the convergence condition is not met even after reaching the preset maximum number of iterations, the candidate machining state is determined to be an unstable calculation state, and the calculation is re-performed by changing the robot's redundant posture, the tool axis direction, or reducing the cutting parameters.

[0024] Preferably, in step S5: The dynamic responses of the tool side and the thin-walled structural part side along the normal to the machined surface, the feed direction, and the tool axis direction are obtained respectively; The responses on both sides are transformed to a unified coordinate reference to obtain the relative dynamic response in the corresponding direction; When the relative vibration of the tool axis direction does not reach the dynamic contact triggering condition, the dynamic stability is evaluated by using the cutting contact area under the cutting balance state. When the relative vibration in the direction of the tool axis reaches the dynamic contact trigger condition, the axial cutting area is divided into a relatively stable contact area that continuously participates in cutting and a dynamic contact area that periodically enters or exits the material due to relative vibration, and the dynamic cutting load is re-determined based on the dynamic contact area. The dynamic stability of the corresponding candidate machining state is determined again based on the corrected dynamic cutting load.

[0025] Preferably, the allowable feed range corresponding to each candidate processing state is determined in the following manner: The motion state of each joint of the robot as the feed speed changes is determined based on the robot's current candidate posture and machining direction, and the first feed range under the robot's motion constraints is determined based on the robot's joint position, allowable speed and allowable acceleration. Starting from the preset minimum evaluation feed, the candidate feed rate is gradually increased within the first feed range, and the cutting load and three-dimensional dynamic response are re-predicted for each candidate feed rate; When increasing the candidate feed rate causes the predicted vibration response in any direction to exceed the corresponding allowable value, stop increasing the feed rate and determine the feed rate that still meets the vibration requirements as the allowable feed limit for the current candidate machining state. The allowable feed range is determined based on the preset minimum stable machining feed and the allowable feed upper limit.

[0026] Preferably, when screening candidate machining states in step S6, contour error, dynamic stability, no interference between tool and workpiece, and robot motion constraints are first used as hard constraints, and candidate machining states that do not meet any hard constraints are deleted. For candidate machining states that satisfy all hard constraints, the machining state with smaller relative deformation in the predicted normal direction is preferred. When the predicted normal relative deformation of multiple candidate machining states is within the same preset accuracy range, the machining state with a larger dynamic stability margin is selected first. When the dynamic stability margins of multiple candidate processing states all meet the same preset stability range, the processing state with the larger allowable feed upper limit is selected first. When multiple candidate machining states still exist, select the machining state with smaller changes in robot joints and tool axis direction compared to adjacent machining segments.

[0027] Preferably, in step S7, at least three representative feature positions are selected within the effective contact area between the tool and the thin-walled structural member, distributed along the tool axis direction. For side-cutting, the representative feature locations include feature locations at different heights within the effective axial cutting range; For ball end mill machining, the representative feature locations are distributed within the contact zone where the spherical surface actually participates in the cutting; The overall profile offset is determined based on the relative deformation of the normal at each representative feature location; The relative tilting trend between the tool and the thin-walled structural component is determined based on the variation trend of the normal relative deformation along the tool axis at each representative feature position. Move the tool position in a direction that counteracts the overall contour offset, and adjust the tool axis direction in a direction that counteracts the relative tilting trend.

[0028] Preferably, in step S8, after each modification of the tool position or tool axis direction: Re-establish the tool-workpiece contact area based on the modified tool position and tool axis direction; Re-execute the cutting load and quasi-static relative deformation iteration; Reassess the three-dimensional dynamic stability and robot motion state; If the predicted residual contour error still exceeds the corresponding machining accuracy requirement, continue to correct the tool position or tool axis direction. When the modified tool axis direction exceeds the non-interference range, the corresponding machining state cannot achieve the cutting balance state, or the dynamic stability does not meet the requirements, the current compensation state is abandoned and a new machining state is selected. The five-axis toolpath is output only after the requirements for residual contour error, dynamic stability, interference-free state and robot motion continuity are met.

[0029] Preferably, the data correspondence in step S9 is established in the following manner: Set a machining status indicator for each machining segment and its corresponding tool position range; The robot control system outputs the current processing status indicator and corresponding time information; The actual cutting load, robot joint status, and vibration response are categorized into the data windows corresponding to the respective machining status identifiers according to time information. The actual contour, actual wall thickness, or actual material after processing is removed from the boundary and mapped to the corresponding processing segment according to the spatial position. The model state of the machining segment and related subsequent machining segments is corrected only by using cutting load, vibration, robot motion state and machining result data that correspond to the same machining segment.

[0030] Preferably, the deviation source determination in step S9 includes: When the actual cutting load continues to deviate from the predicted cutting load, and the actual machining profile, actual material removal boundary, and dynamic response do not show deviations that meet the corresponding reconstruction conditions, the predicted cutting load state is corrected first. When the main response frequency or vibration amplitude in the actual dynamic response continues to deviate from the predicted dynamic response, and the actual material removal boundary does not show a deviation that meets the corresponding reconstruction conditions, the local dynamic response state is corrected first. When the actual machining contour, actual wall thickness, or actual material removal boundary reaches the material state reconstruction condition, the material removal state model is updated, and the updated material removal state propagates to the subsequent local quasi-static compliance state, local dynamic response state, and tool-workpiece contact state. When at least two types of deviations among the actual material removal state deviation, cutting load deviation, and dynamic response deviation have a continuous correspondence and simultaneously meet their respective reconstruction conditions, multi-state joint reconstruction is performed on the relevant subsequent processing sections.

[0031] Preferably, the multi-state joint reconstruction includes: Update the material removal state model with the actual processing contour, actual wall thickness, or actual material removal boundary obtained from actual measurements; The remaining wall thickness, local boundary constraints, and remaining material distribution of the subsequent processing section are re-determined based on the updated material removal status. Regain the local quasi-static flexibility state of the subsequent processing section and correct the local dynamic response state according to the actual dynamic response; Correct the predicted cutting load state using the actual cutting load; The tool-workpiece contact area corresponding to each candidate tool axis direction is redefined based on the updated actual remaining material boundary. The candidate robot redundant posture screening, cutting balance iteration, dynamic stability judgment, allowable feed range determination, posture advance migration planning, and tool position and tool axis direction compensation are re-executed.

[0032] Compared with existing technologies, the five-axis machining method for thin-walled structural parts of this invention models the quasi-static compliance and dynamic response of the thin-walled structural parts separately, and uses them for contour error prediction and machining stability judgment, avoiding the confusion caused by describing different physical responses with a single stiffness parameter. Considering the reverse influence of machining deformation on the actual tool-workpiece contact state and cutting load, the cutting equilibrium state is obtained by repeatedly updating the contact area, cutting load, and quasi-static relative deformation, ensuring that the deformation prediction on the robot side and the thin-walled part side is consistent with the actual cutting contact relationship. After compensating for the tool position and tool axis direction, the tool-workpiece contact area is re-established, and deformation and stability evaluation are re-executed, avoiding the invalidation of the prediction results before compensation due to changes in the cutting state caused by the compensation itself. Simultaneously considering the relative dynamic response of the machined surface normal, feed direction, and tool axis direction, the method can reconstruct the dynamic contact area when relative vibration in the tool axis direction changes the axial contact range, improving the rationality of dynamic stability judgment under weak stiffness robot machining conditions. The allowable feed range is determined by combining the robot's motion capabilities and the permissible dynamic response of the thin-walled structure, enabling different material removal stages to adopt feed rates adapted to the current structural state while ensuring machining accuracy and stability. The stiffness drop region is predicted based on the future material removal state, and an advance attitude migration range is determined upstream of this region. This allows the robot to complete major attitude adjustments in areas where the workpiece still has high stiffness, reducing machining state disturbances caused by significant changes in robot attitude after entering low-stiffness regions. Online cutting loads, vibration states, and actual machining results are correlated according to machining segments and time information to avoid data from different spatial locations or machining stages being incorrectly used for the same model correction. A bias causal gating update mechanism is adopted, selecting the model to be corrected based on the different sources of cutting load deviation, dynamic state deviation, and actual material state deviation. Joint reconstruction is only performed when the basic material state changes or multiple related states deviate together, thereby reducing the propagation of erroneous states and unnecessary full model updates. The actual material removal state serves as the structural basis for subsequent model reconstruction. When the actual material state changes, the process is gradually re-transmitted from the material state to the quasi-static compliance, dynamic response, cutting area, robot posture, tool axis posture, feed range, and compensation toolpath, so that subsequent robot five-axis machining is always based on the current actual workpiece state. [Attached Image Description] Figure 1 This is the overall flowchart of the five-axis machining method for thin-walled structural parts of the robot according to the present invention.

Detailed Implementation Methods

[0033] It should be noted that in the embodiments of the present invention, all directional indications (such as up, down, left, right, front, back, etc.) are limited to relative positions on the specified view, rather than absolute positions.

[0034] Furthermore, in this invention, descriptions involving "first," "second," etc., are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0035] Please see Figure 1 The present invention provides a five-axis machining method for thin-walled structural components using an industrial robot with at least six degrees of freedom, carrying a machining spindle and milling cutters to perform five-axis linkage machining on the thin-walled structural components. The method includes the following steps: Step S1: Obtain the initial geometric model, target machining profile, blank state, clamping constraints and machining allowance of the thin-walled structural part to be machined, establish a material removal state model that characterizes the spatial distribution of the current remaining material, and divide the predetermined five-axis toolpath into multiple machining segments to be executed sequentially based on the curvature change of the surface to be machined, the change of the remaining wall thickness and the change of the theoretical tool axis direction. Step S2: Before executing the current processing segment, determine the remaining wall thickness, local boundary constraints, and remaining material distribution of the current processing segment based on the material removal state model, and establish a local quasi-static compliance state for contour error prediction and a local dynamic response state for vibration stability judgment, respectively; at the same time, predict the local dynamic response state after material removal of the subsequent processing segments according to the predetermined processing sequence; when the predicted local stiffness drop between adjacent subsequent processing segments reaches the preset sudden drop condition, determine the stiffness drop region and the attitude advance migration interval located upstream of the stiffness drop region; Step S3: Under the constraints of no interference between the tool and the workpiece, the range of motion of the robot joints, the robot's singular configuration avoidance, and the attitude continuity, the tool's forward tilt angle, the tool's side tilt angle, and the redundant attitude of the robot around the tool axis are changed to form multiple candidate machining states. For each candidate machining state, the rigid geometric contact area between the tool and the thin-walled structure is re-determined according to the current material removal state, and the effective cutting edge position, material removal state, and cutting speed state are determined according to the rigid geometric contact area to obtain the corresponding initial predicted cutting load. Step S4: Determine the quasi-static displacement of the tool side based on the initial predicted cutting load, the compliance state corresponding to the robot's current candidate posture, and the compliance state of the machining spindle and milling tool. Determine the quasi-static displacement of the thin-walled structure side based on the local quasi-static compliance state of the current machining segment. Transform the quasi-static displacement of the tool side and the quasi-static displacement of the thin-walled structure side to the same workpiece coordinate system to determine the quasi-static relative deformation between the tool and the thin-walled structure. Correct the actual spatial relationship between the tool and the thin-walled structure based on the quasi-static relative deformation, redetermine the cutting contact area and cutting load, and recalculate the quasi-static relative deformation until the changes in cutting load and normal relative deformation in two adjacent calculations both meet the preset convergence conditions. Determine the converged cutting contact area, cutting load, and quasi-static relative deformation as the cutting equilibrium state of the corresponding candidate machining state. Step S5: Based on the cutting balance state, determine the dynamic responses of the tool side and the thin-walled structural part side along the normal direction of the machined surface, the feed direction, and the tool axis direction, respectively. Judge the machining stability based on the relative dynamic responses of the two. When the relative vibration in the tool axis direction reaches a level that can change the actual contact range of the tool axis, redetermine the dynamic cutting contact area affected by the vibration, and recalculate the dynamic cutting load and dynamic stability state based on the changed actual contact state. At the same time, determine the allowable feed range corresponding to each candidate machining state based on the robot joint motion capability and the allowable vibration state of the thin-walled structural part. Step S6: Eliminate candidate machining states that do not meet the contour error requirements, dynamic stability requirements, no interference requirements, robot motion constraints, or have no effective allowable feed range; screen the remaining candidate machining states in the order of machining accuracy, dynamic stability margin, allowable feed capacity, and attitude continuity of adjacent machining segments, and plan the machining state sequence for multiple consecutive machining segments; when the attitude early migration interval exists, allocate the redundant robot attitude and tool axis attitude change required after entering the stiffness drop region to the consecutive machining segments within the attitude early migration interval, so that the robot completes at least the main part of the attitude transformation before entering the stiffness drop region; Step S7: Based on the determined target machining state, compensate the theoretical five-axis toolpath. Specifically, determine the tool position compensation amount based on the normal relative deformation of multiple representative feature positions within the effective cutting area, determine the tool axis direction compensation amount based on the changing trend of the normal relative deformation between each representative feature position along the tool axis direction, and determine the target feed state within the allowable feed range to form a compensated toolpath to be verified that includes tool position, tool axis direction, and time-varying feed information. Step S8: For the toolpath to be checked and compensated, redetermine the compensated tool-workpiece contact area, cutting load, quasi-static relative deformation, and dynamic stability. If the residual contour error, dynamic stability, tool-workpiece non-interference state, and robot motion continuity do not simultaneously meet the requirements, correct the tool position, tool axis direction, or machining parameters again, and recheck. When all the check conditions are met simultaneously, determine the corresponding toolpath as the five-axis toolpath for the current machining segment. Step S9: Control the robot to complete the current machining segment according to the five-axis toolpath. During the machining process, simultaneously collect the actual cutting load, robot motion state, and vibration response that correspond to the current machining segment and tool position. After completing the predetermined machining segment, obtain at least one of the following machining results: actual machining contour, actual wall thickness, or actual material removal boundary. Based on the corresponding deviation between the actual data and the predicted data, identify the source of the deviation and selectively correct the cutting load prediction state, local dynamic response state, or material removal state model according to the source of the deviation. When the material removal state is corrected, or multiple related states simultaneously show a continuous deviation that reaches the preset reconstruction condition, redetermine the local quasi-static compliance state, local dynamic response state, tool-workpiece contact area, candidate tool axis direction, robot redundant posture, allowable feed range, and toolpath compensation amount of the subsequent machining segment, starting from the corrected actual material removal state. Step S10: Perform subsequent machining segments according to the updated machining status until the five-axis machining of the thin-walled structural component is completed.

[0036] When dividing the five-axis toolpath into machining segments in step S1: The tool position is obtained sequentially along the theoretical five-axis toolpath; The curvature of the machined surface, the theoretical remaining wall thickness, and the theoretical tool axis direction at each tool position are obtained respectively. When at least one of the surface curvature change, theoretical remaining wall thickness change, and theoretical tool axis direction change between adjacent tool positions meets the corresponding preset segmentation conditions, the machining segment boundary is established. The material removal state, local structural state, and tool-workpiece contact state within the same processing section are kept within a predetermined range of variation.

[0037] In step S2: The local quasi-static compliance state includes at least the unit load displacement response of the current machining region along the normal to the machined surface and the direction of the main cutting load; The local dynamic response state includes at least the dynamic compliance characteristics of the current machining area along the normal of the machined surface, the feed direction and the tool axis direction, at least one low-order natural frequency and the corresponding damping characteristics. Multiple material removal state samples are pre-established based on different remaining wall thicknesses, local boundary constraints, and remaining material distributions, and the local quasi-static compliance state and local dynamic response state corresponding to each material removal state sample are obtained. During the formal processing, the corresponding state sample is selected based on the current material removal state, or the initial state parameters of the current processing area are obtained based on adjacent state samples, and the state parameters corresponding to the subsequent processing sections are corrected using actual processing data.

[0038] Determining the attitude advance migration interval in step S2 includes: Based on the current actual material removal status and the predetermined processing sequence, the predicted remaining material status after the completion of multiple future processing stages is generated sequentially. The predicted local dynamic response state of the corresponding region is determined based on the predicted remaining material state of each region. Compare the local stiffness changes between adjacent future processing sections; When the stiffness decrease or the stiffness decrease rate reaches the preset sudden drop condition, the corresponding processing section will be identified as the stiffness sudden drop region. Search upstream of the stiffness drop region along the processing direction for a processing segment that can meet the requirements of continuous robot posture change and processing stability, and determine the continuous processing segment obtained by the search as the posture advance migration interval.

[0039] Before the actual processing, multiple calibration poses are selected in the robot workspace covering the actual processing area. Known loads in multiple directions are applied to the robot end effector and the pose changes of the robot end effector are measured to obtain the equivalent stiffness parameters of the robot joints. For each candidate machining state, the robot end effector flexibility is obtained based on the current joint position of the robot and the equivalent stiffness parameters of the robot joint, and the robot end effector flexibility is converted to the tool-workpiece contact area; The equivalent compliance of the machining spindle and milling cutter is added to the converted robot end effector compliance to determine the overall compliance state of the tool side. The tool side quasi-static displacement includes at least the translation of the robot end effector, the additional displacement of the tool position caused by the rotation of the robot end effector, the elastic deformation of the machining spindle, and the bending deformation of the milling tool.

[0040] In step S3, the process of redetermining the tool-workpiece contact area for each candidate tool axis direction includes: Determine the boundary of the remaining material entity based on the current material removal state model; Determine the position where the cutting edge enters the remaining material, the position where it leaves the remaining material, and the effective cutting range along the tool axis in the current candidate tool axis direction; For ball end mills, the actual spherical area involved in cutting is determined based on the positional relationship between the spherical cutting edge and the remaining material entity. The corresponding cutting speed state and instantaneous material removal state are determined based on the actual position of the cutting edge involved in the cutting. The cutting load is determined separately for different candidate tool axis directions, and the same cutting load is not directly used for deformation evaluation in different candidate tool axis directions.

[0041] The cutting balance state in step S4 is determined in the following manner: The cutting load obtained based on the rigid geometric contact region is used as the first predicted cutting load; The quasi-static displacement of the tool side and the quasi-static displacement of the thin-walled structural component side are determined by the first predicted cutting load. The predicted actual position of the tool relative to the thin-walled structure is changed based on the quasi-static relative deformation of the two, and the tool-workpiece contact area and predicted cutting load for the next cutting are redefined. Repeatedly perform the update of the contact area, the update of the cutting load, and the update of the quasi-static relative deformation; The iteration ends when the representative cutting load change and normal relative deformation change obtained in two consecutive calculations are both less than the corresponding convergence allowable value. If the convergence condition is not met even after reaching the preset maximum number of iterations, the candidate machining state is determined to be an unstable calculation state, and the calculation is re-performed by changing the robot's redundant posture, the tool axis direction, or reducing the cutting parameters.

[0042] In step S5: The dynamic responses of the tool side and the thin-walled structural part side along the normal to the machined surface, the feed direction, and the tool axis direction are obtained respectively; The responses on both sides are transformed to a unified coordinate reference to obtain the relative dynamic response in the corresponding direction; When the relative vibration of the tool axis direction does not reach the dynamic contact triggering condition, the dynamic stability is evaluated by using the cutting contact area under the cutting balance state. When the relative vibration in the direction of the tool axis reaches the dynamic contact trigger condition, the axial cutting area is divided into a relatively stable contact area that continuously participates in cutting and a dynamic contact area that periodically enters or exits the material due to relative vibration, and the dynamic cutting load is re-determined based on the dynamic contact area. The dynamic stability of the corresponding candidate machining state is determined again based on the corrected dynamic cutting load.

[0043] The allowable feed range for each candidate processing state is determined as follows: The motion state of each joint of the robot as the feed speed changes is determined based on the robot's current candidate posture and machining direction, and the first feed range under the robot's motion constraints is determined based on the robot's joint position, allowable speed and allowable acceleration. Starting from the preset minimum evaluation feed, the candidate feed rate is gradually increased within the first feed range, and the cutting load and three-dimensional dynamic response are re-predicted for each candidate feed rate; When increasing the candidate feed rate causes the predicted vibration response in any direction to exceed the corresponding allowable value, stop increasing the feed rate and determine the feed rate that still meets the vibration requirements as the allowable feed limit for the current candidate machining state. The allowable feed range is determined based on the preset minimum stable machining feed and the allowable feed upper limit.

[0044] When screening candidate machining states in step S6, contour error, dynamic stability, no interference between tool and workpiece, and robot motion constraints are first used as hard constraints, and candidate machining states that do not meet any hard constraints are deleted. For candidate machining states that satisfy all hard constraints, the machining state with smaller relative deformation in the predicted normal direction is preferred. When the predicted normal relative deformation of multiple candidate machining states is within the same preset accuracy range, the machining state with a larger dynamic stability margin is selected first. When the dynamic stability margins of multiple candidate processing states all meet the same preset stability range, the processing state with the larger allowable feed upper limit is selected first. When multiple candidate machining states still exist, select the machining state with smaller changes in robot joints and tool axis direction compared to adjacent machining segments.

[0045] In step S7, at least three representative feature positions are selected from the distribution range along the tool axis within the effective contact area between the tool and the thin-walled structural component. For side-cutting, the representative feature locations include feature locations at different heights within the effective axial cutting range; For ball end mill machining, the representative feature locations are distributed within the contact zone where the spherical surface actually participates in the cutting; The overall profile offset is determined based on the relative deformation of the normal at each representative feature location; The relative tilting trend between the tool and the thin-walled structural component is determined based on the variation trend of the normal relative deformation along the tool axis at each representative feature position. Move the tool position in a direction that counteracts the overall contour offset, and adjust the tool axis direction in a direction that counteracts the relative tilting trend.

[0046] In step S8, after each modification of the tool position or tool axis direction: Re-establish the tool-workpiece contact area based on the modified tool position and tool axis direction; Re-execute the cutting load and quasi-static relative deformation iteration; Reassess the three-dimensional dynamic stability and robot motion state; If the predicted residual contour error still exceeds the corresponding machining accuracy requirement, continue to correct the tool position or tool axis direction. When the modified tool axis direction exceeds the non-interference range, the corresponding machining state cannot achieve the cutting balance state, or the dynamic stability does not meet the requirements, the current compensation state is abandoned and a new machining state is selected. The five-axis toolpath is output only after the requirements for residual contour error, dynamic stability, interference-free state and robot motion continuity are met.

[0047] The data correspondence in step S9 is established in the following manner: Set a machining status indicator for each machining segment and its corresponding tool position range; The robot control system outputs the current processing status indicator and corresponding time information; The actual cutting load, robot joint status, and vibration response are categorized into the data windows corresponding to the respective machining status identifiers according to time information. The actual contour, actual wall thickness, or actual material after processing is removed from the boundary and mapped to the corresponding processing segment according to the spatial position. The model state of the machining segment and related subsequent machining segments is corrected only by using cutting load, vibration, robot motion state and machining result data that correspond to the same machining segment.

[0048] The deviation source identification in step S9 includes: When the actual cutting load continues to deviate from the predicted cutting load, and the actual machining profile, actual material removal boundary, and dynamic response do not show deviations that meet the corresponding reconstruction conditions, the predicted cutting load state is corrected first. When the main response frequency or vibration amplitude in the actual dynamic response continues to deviate from the predicted dynamic response, and the actual material removal boundary does not show a deviation that meets the corresponding reconstruction conditions, the local dynamic response state is corrected first. When the actual machining contour, actual wall thickness, or actual material removal boundary reaches the material state reconstruction condition, the material removal state model is updated, and the updated material removal state propagates to the subsequent local quasi-static compliance state, local dynamic response state, and tool-workpiece contact state. When at least two types of deviations among the actual material removal state deviation, cutting load deviation, and dynamic response deviation have a continuous correspondence and simultaneously meet their respective reconstruction conditions, multi-state joint reconstruction is performed on the relevant subsequent processing sections.

[0049] The multi-state joint reconstruction includes: Update the material removal state model with the actual processing contour, actual wall thickness, or actual material removal boundary obtained from actual measurements; The remaining wall thickness, local boundary constraints, and remaining material distribution of the subsequent processing section are re-determined based on the updated material removal status. Regain the local quasi-static flexibility state of the subsequent processing section and correct the local dynamic response state according to the actual dynamic response; Correct the predicted cutting load state using the actual cutting load; The tool-workpiece contact area corresponding to each candidate tool axis direction is redefined based on the updated actual remaining material boundary. The candidate robot redundant posture screening, cutting balance iteration, dynamic stability judgment, allowable feed range determination, posture advance migration planning, and tool position and tool axis direction compensation are re-executed.

[0050] Specifically, the present invention aims to provide a five-axis machining method for thin-walled structural parts of a robot, in order to reduce the contour error and machining vibration caused by the combined flexibility of the robot side and the thin-walled workpiece side, improve the consistency between the robot five-axis machining model and the actual machining state under continuous material removal conditions, and improve the adverse effects of robot posture changes on machining stability near low stiffness regions.

[0051] This invention employs an industrial robot with at least six degrees of freedom to carry a machining spindle and milling cutters for machining.

[0052] The five-axis linkage machining described in this invention refers to the coordinated control of three positional degrees of freedom at the tool center point and two attitude degrees of freedom along the tool axis during machining. The robot can utilize the remaining attitude degrees of freedom around the tool axis to change its joint configuration and improve its machining performance while maintaining the target tool position geometry.

[0053] 1. Establishment of material removal status The material to be removed is determined based on the blank model of the thin-walled structural part to be processed and the target part model.

[0054] The material removal state model is used to record the actual or predicted remaining material state of the workpiece at the current processing moment, and includes at least: the current remaining material boundary; the current processed area; the current unprocessed allowance; the local remaining wall thickness; and the boundary constraints between the workpiece and the clamping structure.

[0055] During initial processing, the theoretical blank state is used to establish the first material removal state.

[0056] When the actual contour, wall thickness or material boundary measurement results are obtained after processing, it is determined whether the theoretical material removal state is consistent with the actual state based on the measurement results.

[0057] Only when the actual material state deviation reaches the corresponding reconstruction condition will the actual measurement results be used to replace the corresponding theoretical material boundary, so as to avoid frequent model changes caused by measurement noise.

[0058] 2. Processing Section Division Continuous toolpaths are obtained based on theoretical five-axis toolpaths.

[0059] The surface curvature, theoretical remaining wall thickness, and theoretical tool axis direction variation of continuous tool positions are compared.

[0060] A new processing segment is established when any of the above state changes meets the corresponding segmentation conditions.

[0061] The segmentation conditions can be determined based on the target machining accuracy, workpiece size, and computing power. Shorter machining segments are used in finishing or low-stiffness areas, while longer machining segments are used in areas with high stiffness and gentle curvature changes.

[0062] The processing section serves as the basic spatial unit for subsequent model prediction, online data correspondence, and state updates.

[0063] 3. Separation of quasi-static compliance state and dynamic response state Quasi-static flexibility state and dynamic response state are established for thin-walled structural components.

[0064] Quasi-static compliance states are used to predict the elastic displacement of the workpiece based on stable cutting loads, and further predict machining profile deviations.

[0065] Dynamic response status is used to evaluate the vibration response and stability of thin-walled structural components under periodic cutting excitation.

[0066] Before formal processing, multiple material removal state samples are established based on different remaining wall thicknesses, different local boundary conditions, and different remaining material distributions.

[0067] For each state sample, a quasi-static load in a known direction is applied to obtain the displacement response corresponding to a unit load.

[0068] At the same time, at least one low-order dynamic response feature is obtained for each state sample, including natural frequency, damping and dynamic compliance changes in the corresponding direction.

[0069] Establish a correspondence between the material removal state and the quasi-static compliance state and dynamic response state.

[0070] During actual processing, the initial model state is obtained based on the current material removal state.

[0071] 4. Future Stiffness State Prediction For the predetermined number of processing segments following the current processing segment, simulate subsequent material removal based on the current material removal status and the theoretical processing sequence.

[0072] The predicted state of the remaining material after the completion of each future processing segment is generated sequentially.

[0073] The predicted local stiffness state of the future processing area is determined based on the predicted state of the remaining material.

[0074] Compare adjacent predicted states in turn.

[0075] When the local stiffness decreases to the preset sudden drop condition, the corresponding area is marked as the stiffness sudden drop area.

[0076] The preset sudden drop conditions are not limited to a fixed, uniform value.

[0077] In actual implementation, the stiffness variation range of adjacent processing sections under normal continuous material removal conditions is obtained during the offline processing verification stage, and stiffness changes that significantly exceed the normal variation range and cause significant deterioration of the expected profile error or vibration state are used as the basis for judging sudden drops.

[0078] 5. Determining the advance attitude transition range After identifying the region where stiffness suddenly drops, search for the attitude transition region upstream of that region along the machining direction.

[0079] The posture transition region must simultaneously meet the following requirements: the current workpiece stiffness is higher than the stiffness drop region; the required robot joint posture changes can be performed; the tool and workpiece can still remain free from interference after the posture change; and the predicted machining error and dynamic response during the posture transition process both meet the requirements of the current machining stage.

[0080] The continuous processing segments that meet the above conditions are taken as the attitude advance migration intervals.

[0081] After the robot enters the pre-adjustment range, it changes redundant postures and necessary tool axis postures for each processing segment, so that the main posture changes are completed before entering the stiffness drop region.

[0082] 6. Establishing the robot's flexibility state Select multiple calibration poses within the robot's designated workspace.

[0083] Multiple known loads are applied to the robot's end effector in multiple directions, and the corresponding end effector displacement or pose changes are measured.

[0084] The equivalent stiffness of the robot joints is determined by combining the joint angles of the robot at each calibration pose.

[0085] During the actual processing, the current joint state is obtained based on the candidate robot's posture, and the corresponding robot end effector flexibility is determined.

[0086] After the robot end effector undergoes a slight rotation, due to the spatial distance between the robot flange and the tool tip, this rotation will produce additional displacement at the tool's cutting point.

[0087] Therefore, the additional displacement of the tool caused by the translation and rotation of the robot end effector, the machining spindle compliance, and the tool compliance are all included in the quasi-static deformation of the tool side.

[0088] 7. Generation of candidate processing states For the current machining section, the allowable tool axis adjustment space is determined based on the theoretical tool axis direction.

[0089] Change the tool rake angle and tool side tilt angle respectively.

[0090] For each candidate tool axis direction, the redundant posture of the robot around the tool axis direction is further changed.

[0091] Perform the following checks on all candidate machining states: tool-workpiece interference check; robot-workpiece collision check; robot self-collision check; robot joint position constraint check; robot joint motion capability check; robot singular configuration check. Delete machining states that do not satisfy any of these constraints.

[0092] 8. Initial cutting contact state between the tool and the workpiece For each remaining candidate processing state, the boundary of the remaining material in the workpiece is determined using the current material removal state model.

[0093] Determine the spatial position of the tool relative to the remaining material based on the current candidate tool axis direction.

[0094] Determine the position where the cutting edge enters the material, the position where it leaves the material, and the area of ​​the cutting edge that effectively participates in the cutting.

[0095] For ball end mill machining, determine the contact zone on the spherical surface that actually participates in cutting.

[0096] For side-cutting machining, determine the actual cutting height range along the tool axis.

[0097] The effective cutting speed and material removal status are determined based on the actual cutting position.

[0098] The first predicted cutting load is obtained based on tool geometry, workpiece material, spindle speed, feed state, axial depth of cut, radial cutting state, and actual contact range.

[0099] Before formal machining, it is preferable to perform trial cutting with the same material or a material with equivalent cutting properties to calibrate the predicted cutting load parameters.

[0100] 9. Cutting load—relative deformation—iterative contact area The first prediction is that the cutting load will act on both the tool side and the thin-walled structural component side.

[0101] The lateral static displacement of the tool is determined based on the robot, machining spindle, and tool flexibility.

[0102] The quasi-static displacement of the workpiece side is determined based on the local quasi-static flexibility of the current thin-walled structural component.

[0103] Both are converted to the workpiece coordinate system.

[0104] Calculate the quasi-static relative deformation between the tool side and the workpiece side at the corresponding cutting contact position.

[0105] The actual spatial relationship between the tool and the workpiece is predicted based on the quasi-static relative deformation change.

[0106] The tool-workpiece contact area is recalculated based on the changed spatial relationship.

[0107] Then recalculate the cutting load based on the new contact area.

[0108] Repeat the following steps: cut contact area calculation; cutting load prediction; tool side and workpiece side deformation calculation; relative deformation calculation; actual contact relationship correction.

[0109] After each iteration, the representative cutting load and normal relative deformation of the current iteration and the previous iteration are compared respectively.

[0110] Once both changes fall within the corresponding convergence allowable range, the current candidate state is considered to have reached the cutting equilibrium state.

[0111] The convergence allowable range is determined based on the repeatability of cutting load measurement, the allowable error of machining dimensions, and the accuracy of numerical calculation.

[0112] To ensure the computation process can be completed, a maximum number of iterations is preset.

[0113] If convergence is not achieved after reaching the maximum number of iterations, the non-converged calculation results are not used directly. Instead, the current candidate state is deleted or the feed and cutting depth are reduced before recalculation.

[0114] 10. Three-dimensional relative dynamic response For candidate machining states that have reached cutting equilibrium, the dynamic responses of the tool side and workpiece side along the normal to the machined surface, the feed direction, and the tool axis direction are obtained respectively.

[0115] The tool-side dynamic response is obtained based on the current robot posture, the dynamic state of the spindle tool assembly, and periodic cutting excitation.

[0116] The workpiece-side dynamic response is obtained based on the local dynamic response state corresponding to the current material removal state.

[0117] The two coordinates are converted to a unified workpiece coordinate datum and projected onto the corresponding feature direction.

[0118] Normal dynamic response is mainly used to determine vibrations affecting the machining profile; The dynamic response in the feed direction is used to determine the relative vibration along the toolpath direction; The dynamic response of the tool axis direction is used to determine whether the axial contact state has undergone periodic changes.

[0119] 11. Dynamic Contact Area Determination When the predicted relative vibration in the tool axis direction is less than the dynamic contact triggering condition, it is assumed that the axial tool-workpiece contact range will not change significantly due to vibration, and the dynamic stability is judged by the cutting contact area under the cutting equilibrium state.

[0120] When the relative vibration along the tool axis reaches the dynamic contact triggering condition, the periodic displacement range of the tool relative to the workpiece along the tool axis is determined based on the predicted vibration range.

[0121] The entire axial contact area is divided into: a relatively stable contact area that continuously participates in cutting within the corresponding vibration cycle; and a dynamic contact area that periodically enters or exits the material due to the relative axial movement of the tool and the workpiece.

[0122] The cyclic cutting load state was redefined based on the two regions, and the three-dimensional dynamic response was re-evaluated.

[0123] The dynamic contact triggering condition is determined based on whether the axial relative vibration can cause the actual contact boundary to move to a degree that affects the accuracy of the cutting load calculation.

[0124] 12. Allowable feed interval For each candidate processing state that meets the stability requirements, determine the allowable feed range.

[0125] First, based on the robot's current joint position and theoretical toolpath, the feed rate to be evaluated is gradually increased, and the required speed and acceleration for each joint of the robot are determined.

[0126] The allowable range of robot motion is defined as the feed range that will not cause any robot joint to exceed its permissible motion capability.

[0127] Then, the feed rate is gradually increased starting from the preset minimum stable feed.

[0128] For each increase in feed rate, the cutting load and dynamic response are recalculated based on the new feed conditions.

[0129] When the predicted vibration first exceeds the workpiece's allowable vibration level, the feed rate should be stopped.

[0130] The previous feed state that still meets the dynamic requirements is taken as the upper limit of the allowable feed for the current candidate posture.

[0131] The minimum stable feed and the upper limit of the allowable feed are used to form the allowable feed range for the current candidate state.

[0132] If no valid allowable feed range exists, delete the current candidate state or change the tool axis direction, robot redundant posture, and cutting depth before recalculating.

[0133] 13. Processing status stratification screening This invention does not rely on a single comprehensive evaluation function with uncertain weights.

[0134] First, filter according to hard constraints.

[0135] Candidate states must simultaneously meet the following requirements: predicted contour error; dynamic stability; no interference; and robot joint motion capability.

[0136] Only processing states that simultaneously meet the above conditions will proceed to the next screening stage.

[0137] Then, the predicted normal relative deformation is compared first.

[0138] Prioritize those with smaller relative deformation in the predicted normal direction.

[0139] When multiple states fall within the same allowable accuracy range, the dynamic stability margin is then compared.

[0140] Those with larger dynamic stability margins will be given priority.

[0141] If multiple states are still at the same stability level, then compare the allowed feed limits.

[0142] Those with higher feed limits are given priority.

[0143] Finally, by comparing the changes in robot joints and tool axis between the preceding and following processing stages, a state that can form a more continuous processing motion is selected.

[0144] 14. Processing status planning across processing sections The above candidate state screening will be performed on multiple consecutive processing sections in the future.

[0145] Record the preferred processing state of each processing segment and the attitude transition between adjacent processing segments.

[0146] For ordinary regions where there is no sudden drop in stiffness, the changes in robot joints and tool axes in adjacent processing sections are kept continuous.

[0147] If the situation is expected to enter the region of sudden drop in stiffness, the aforementioned attitude is used to advance the migration range.

[0148] The redundant posture of the target robot required to enter the region is determined by optimizing the processing state in the region of sudden stiffness drop.

[0149] The attitude change is assigned segment by segment starting from the beginning of the attitude advance migration interval.

[0150] Each processing segment is only responsible for the attitude change that meets the requirements of joint speed, acceleration, and current processing error.

[0151] This allows the robot to complete its main posture adjustment in a relatively high stiffness region and enter a low stiffness region with a relatively stable target posture.

[0152] 15. Tool position and tool axis direction compensation For the target machining state, select at least three representative feature positions along the tool axis distribution range within the effective tool-workpiece contact area.

[0153] For side-cutting machining, different height positions within the effective cutting height range can be selected.

[0154] For ball end mill machining, multiple representative positions are selected within the actual cutting contact zone of the spherical surface.

[0155] The normal quasi-static relative deformation of each representative position is determined separately.

[0156] When multiple representative positions exhibit a basically consistent contour offset trend, the tool position compensation is determined based on the relative deformation of their representative normals.

[0157] When there is a significant change in the relative normal deformation of each representative position along the tool axis, it is considered that there is a relative tilting error trend.

[0158] Adjust the tool axis direction according to the changing trend so that the adjustment direction offsets the predicted relative tilt.

[0159] Both the tool position change and the tool axis direction change adopt the reverse compensation principle.

[0160] The maximum allowable position compensation increment and attitude compensation increment are set for adjacent tool positions. Compensation changes exceeding the allowable increment are distributed to multiple consecutive tool positions to avoid abrupt changes in robot motion.

[0161] 16. Secondary verification after compensation Since changes in the tool position or tool axis direction will alter the tool-workpiece contact area, the compensated toolpath cannot be directly used as the execution toolpath after the first compensation toolpath is obtained.

[0162] Re-establish the contact relationship between the tool and the current material state for the modified tool position and tool axis.

[0163] Re-execute the cutting area-cutting load-relative deformation iteration.

[0164] A new cutting balance state is obtained.

[0165] Recalculate the three-dimensional dynamic response.

[0166] Reassess: residual contour error after compensation; dynamic stability; no interference between the tool and the workpiece; robot joint motion capability; continuity of motion between adjacent tool positions.

[0167] If any item fails to meet the conditions, continue to modify the compensation amount or reselect the processing status.

[0168] The five-axis toolpath will only be output and executed after all conditions are met.

[0169] This avoids the compensation operation itself causing the original processing status evaluation to become invalid.

[0170] 17. Time-varying feed planning The target feed is determined within the allowable feed range corresponding to each target processing state.

[0171] Under the premise of meeting the requirements of processing accuracy and stability, the feed state that can improve the material removal efficiency should be given priority.

[0172] If the local structural state within the same processing section still changes significantly, the processing section is further divided into multiple feed sub-regions.

[0173] A continuous transition interval is set between adjacent feed sub-regions.

[0174] The feed rate is gradually changed within the transition range to ensure that the robot's joint speed and acceleration meet the motion capability requirements.

[0175] The final five-axis toolpath execution includes: tool position information; tool axis attitude information; robot redundant attitude information; and target time-varying feed information.

[0176] 18. Synchronous acquisition of processing status After the robot starts processing, the control system continuously outputs the current processing segment identifier, tool position range, and time information.

[0177] The cutting force data is correlated with the current machining state according to the time information.

[0178] The robot's joint positions, speeds, and control states are correlated with the corresponding time information.

[0179] Vibration response was also recorded based on a unified time reference.

[0180] For high-speed cutting force and vibration data, representative data for state judgment is extracted within a stable cutting time window.

[0181] Unstable data when the tool first enters or leaves the material, or when the robot performs a rapid posture transition, are marked separately and are not used as the basis for updating the normal machining state.

[0182] After the processing section is completed, the actual contour, wall thickness, or material boundary is obtained.

[0183] The actual measurement results are assigned to the corresponding processing section based on the spatial coordinates.

[0184] Therefore, the cutting load, dynamic response, and final machining result can all correspond to the actual machining position that produces this state.

[0185] 19. Actual relative dynamic response measurement Preferably, a first vibration measuring device is installed at the end of the robot or at the machining spindle, and a second vibration measuring device is installed at the fixture position on the thin-walled structural part or adjacent to the machining area.

[0186] The data acquired by the two measuring devices are converted to a unified workpiece coordinate reference, and the response characteristics of the tool side and workpiece side along the normal, feed direction and tool axis direction are obtained.

[0187] When it is not possible to directly install a vibration measurement device in the machining area of ​​a thin-walled part, a non-contact vibration measurement device can be used to obtain the workpiece-side response.

[0188] Alternatively, the dynamic response of the workpiece side in the current cutting area can be reconstructed based on the calibrated dynamic model on the workpiece side and measurement data from nearby locations.

[0189] The above method obtains the actual relative dynamic response for model updates, rather than directly equating the vibration of a single robot side with the relative vibration of the tool and the workpiece.

[0190] 20. Determining the State Reconstruction Threshold In the stable machining area before or during the initial stage of machining, the normal range of fluctuation of cutting load, vibration amplitude and normal variation of main response frequency under normal conditions is obtained.

[0191] The basic trigger range for cutting load and dynamic state is determined based on the repeatability of sensor measurements and the normal machining fluctuation range.

[0192] The triggering conditions related to machining contour or wall thickness are determined based on the remaining machining allowance and accuracy requirements of the part.

[0193] For material removal boundaries, the trigger condition is set to a boundary deviation sufficient to substantially change the subsequent predicted local compliance, contact area, or remaining machining allowance.

[0194] To avoid instantaneous shocks or measurement noise triggering model reconstruction, it is preferable to consider the state reconstruction condition to be met only when the deviation exists in multiple consecutive stable sampling windows or persists in the stable processing data of a complete processing segment.

[0195] 21. Biased Causal Gating Update This invention does not immediately modify the entire model after detecting a processing deviation.

[0196] When the actual cutting load is consistently higher or lower than the predicted cutting load, while the actual machining boundary and dynamic response are within the normal range, it is first determined to be a deviation in the predicted cutting load state.

[0197] At this point, the subsequent cutting load prediction parameters should be corrected based on the measured cutting load.

[0198] When the main dynamic response frequency or vibration amplitude continues to change, but the actual material removal boundary is still basically consistent with the predicted state, the corresponding local dynamic response state should be corrected first.

[0199] When the actual machining profile, wall thickness, or material removal boundary deviates sufficiently to change the subsequent structural state, the material removal state is updated first.

[0200] The material removal state is the structural foundation state for subsequent processing. Therefore, after updating this state, the local quasi-static flexibility state and dynamic response state of the relevant subsequent processing section are obtained again, and the tool-workpiece contact area is redefined.

[0201] When two or more types of causally related deviations occur simultaneously and continuously, multi-state joint reconstruction is initiated.

[0202] 22. Multi-state joint reconstruction Multi-state joint reconstruction first modifies the material removal state model based on the actual processing results.

[0203] Replace the corresponding theoretical remaining material state with the updated actual remaining material state.

[0204] Redetermine the following aspects for subsequent processing areas: remaining wall thickness; local boundary constraints; remaining material geometry.

[0205] Then the corresponding local quasi-static flexibility state is regained.

[0206] The corresponding dynamic response state is corrected based on the actual vibration results.

[0207] The predicted cutting load state is corrected based on the actual cutting load.

[0208] Since the material state has changed, the tool-workpiece contact area is re-established based on the new material boundary for different candidate tool axis postures.

[0209] Then, the following steps are repeated: robot candidate pose generation; cutting balance iteration; three-dimensional dynamic stability judgment; allowable feed range determination; future stiffness drop prediction; pose advance migration planning; tool position compensation; tool axis direction compensation; and secondary verification after compensation.

[0210] This ensures that when the actual material condition changes, the downstream processing condition is re-established based on the same actual structural foundation.

[0211] 23. Treatment of low-rigidity machining areas When the current or subsequent processing section is in a low-stiffness state, the robot's redundant posture should be changed first.

[0212] By altering the robot's redundant posture around the tool axis, the robot's joint configuration can be changed, thereby improving the robot's stiffness in the main cutting load directions without changing the current tool center point and tool axis direction.

[0213] If the machining requirements still cannot be met, the tool tilt angle or lateral tilt angle can be changed within the allowable range of no interference.

[0214] Since the cutting contact area changes accordingly after the tool axis is changed, the cutting contact area and cutting load need to be recalculated.

[0215] If a stable machining state cannot be achieved after adjusting the posture, the upper limit of the allowable feed rate should be reduced.

[0216] If necessary, further reduce the axial depth of cut or the radial width of cut.

[0217] If any machining parameter changes and affects the cutting load, the cutting load, relative deformation, and dynamic stability calculations must be re-executed.

[0218] Actual processing will only be performed after the calculation results meet the processing requirements again.

[0219] 24. Implementation methods for roughing, semi-finishing, and finishing. This invention can be used for roughing, semi-finishing, and finishing.

[0220] In the roughing stage, priority is given to avoiding the premature formation of thin-walled areas. By planning the processing sequence, materials that can provide auxiliary stiffness for subsequent processing are retained.

[0221] In the semi-finishing stage, local quasi-static deformation prediction and dynamic stability judgment are added based on the current material removal state.

[0222] For regions where a significant drop in stiffness is predicted to occur, necessary finishing allowances should be retained and an early attitude migration range should be set.

[0223] In the finishing stage, the latest actual material removal status is used as the basis for the final processing plan.

[0224] Recalculate the target robot posture, tool axis posture, and allowable feed range.

[0225] The tool position and tool axis direction are compensated together.

[0226] After the compensation state is checked twice, the final contour processing is completed.

[0227] 25. Implementation Method for Ball End Mill Machining When machining free-form surfaces with a ball end mill, changing the tool's rake angle or tilt angle will change the area on the sphere that actually participates in the cutting process.

[0228] Therefore, the effective cutting area of ​​the spherical surface is redefined each time the candidate tool axis direction is changed.

[0229] The appropriate cutting speed and cutting load are determined based on the effective cutting area.

[0230] Then, the different redundant robot postures under the same tool axis direction are compared.

[0231] If a certain tool axis direction can achieve high robot stiffness, but causes the actual effective cutting area to be in an unfavorable cutting position and generate a large cutting load, then that tool axis direction is not considered the preferred state.

[0232] This avoids determining the tool axis posture solely based on the robot's stiffness.

[0233] 26. Side-edge machining implementation method When machining thin-walled curved surfaces using cylindrical end mills, ball nose end mills, or other tools with axial side cutting edges, the actual axial cutting range is divided into multiple representative positions.

[0234] The cutting load state and the relative deformation of the tool side and the workpiece side at the corresponding positions are obtained respectively.

[0235] Position compensation is determined based on the common offset of relative deformation in the normal directions of multiple axial positions.

[0236] The tool axis direction compensation is determined based on the relative deformation differences between multiple axial positions.

[0237] After the first compensation is completed in the tool axis direction, the actual axial cutting range and cutting force are recalculated, and the final execution tool axis is determined through a second verification.

[0238] 27. State Transmission Relationship of the Invention The main states during the processing of this invention are transmitted according to the following relationships: The current actual state of material removal determines the current and future local structural state of the thin-walled structural component; The local structural state determines the quasi-static compliance and dynamic response state, respectively. The tool axis posture and the state of the remaining material together determine the initial tool-workpiece contact area; The cutting contact area determines the cutting load; The cutting load acts simultaneously on both the tool side and the workpiece side, resulting in quasi-static relative deformation; Quasi-static relative deformation reverses the prediction of the actual contact state; The cutting contact area, cutting load, and relative deformation are iteratively combined to form a cutting equilibrium state. The cutting balance state, together with the dynamic responses of the tool side and the workpiece side, jointly determine the three-dimensional relative dynamic state. The three-dimensional relative dynamic state determines the machining stability and the allowable feed range; The future state of material removal determines whether a region of sudden stiffness drop will occur, and whether an early attitude migration range should be set. The target machining state determines the tool position, tool axis direction, and time-varying feed. After tool position and tool axis compensation is completed, the cutting contact area and related machining status are recalculated. Actual processing is only performed after the compensated state has passed a second verification. The actual cutting load, dynamic response, and machining results are correlated according to the same machining segment; The actual deviation is first identified by its source, and then the corresponding status is selectively modified. When the actual material state changes or multiple related states deviate together, a joint reconstruction is performed on the subsequent processing state.

[0239] Therefore, the material removal state, robot posture, tool axis posture, cutting contact area, cutting load, relative deformation, dynamic response, feed state and compensation toolpath in this invention have a clear sequential dependency relationship and are not simply superimposed technical measures that are independent of each other.

[0240] The above description is only a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any modifications, equivalent substitutions and improvements made within the concept of the present invention should be included within the patent protection scope of the present invention.

Claims

1. A five-axis machining method for thin-walled structural components using a robot, comprising employing an industrial robot with at least six degrees of freedom to carry a machining spindle and milling cutters to perform five-axis simultaneous machining of the thin-walled structural components, characterized in that, Includes the following steps: Step S1: Obtain the initial geometric model, target machining profile, blank state, clamping constraints and machining allowance of the thin-walled structural part to be machined, establish a material removal state model that characterizes the current spatial distribution of remaining material, and divide the predetermined five-axis toolpath into multiple sequentially executed machining segments according to the curvature change of the surface to be machined, the change of remaining wall thickness and the change of theoretical tool axis direction. Step S2: Before executing the current processing segment, determine the remaining wall thickness, local boundary constraints, and remaining material distribution of the current processing segment based on the material removal state model, and establish a local quasi-static compliance state for contour error prediction and a local dynamic response state for vibration stability judgment, respectively; at the same time, predict the local dynamic response state after material removal of the subsequent processing segments according to the predetermined processing sequence; when the predicted local stiffness drop between adjacent subsequent processing segments reaches the preset sudden drop condition, determine the stiffness drop region and the attitude advance migration interval located upstream of the stiffness drop region; Step S3: Under the constraints of no interference between the tool and the workpiece, the range of motion of the robot joints, the robot's singular configuration avoidance, and the attitude continuity, the tool's forward tilt angle, the tool's side tilt angle, and the redundant attitude of the robot around the tool axis are changed to form multiple candidate machining states. For each candidate machining state, the rigid geometric contact area between the tool and the thin-walled structure is re-determined according to the current material removal state, and the effective cutting edge position, material removal state, and cutting speed state are determined according to the rigid geometric contact area to obtain the corresponding initial predicted cutting load. Step S4: Determine the quasi-static displacement of the tool side based on the initial predicted cutting load, the compliance state corresponding to the robot's current candidate posture, and the compliance state of the machining spindle and milling tool. Determine the quasi-static displacement of the thin-walled structure side based on the local quasi-static compliance state of the current machining segment. Transform the quasi-static displacement of the tool side and the quasi-static displacement of the thin-walled structure side to the same workpiece coordinate system to determine the quasi-static relative deformation between the tool and the thin-walled structure. Correct the actual spatial relationship between the tool and the thin-walled structure based on the quasi-static relative deformation, redetermine the cutting contact area and cutting load, and recalculate the quasi-static relative deformation until the changes in cutting load and normal relative deformation in two adjacent calculations both meet the preset convergence conditions. Determine the converged cutting contact area, cutting load, and quasi-static relative deformation as the cutting equilibrium state of the corresponding candidate machining state. Step S5: Based on the cutting balance state, determine the dynamic responses of the tool side and the thin-walled structural part side along the normal direction of the machined surface, the feed direction, and the tool axis direction, respectively. Judge the machining stability based on the relative dynamic responses of the two. When the relative vibration in the tool axis direction reaches a level that can change the actual contact range of the tool axis, redetermine the dynamic cutting contact area affected by the vibration, and recalculate the dynamic cutting load and dynamic stability state based on the changed actual contact state. At the same time, determine the allowable feed range corresponding to each candidate machining state based on the robot joint motion capability and the allowable vibration state of the thin-walled structural part. Step S6: Eliminate candidate machining states that do not meet the contour error requirements, dynamic stability requirements, no interference requirements, robot motion constraints, or have no effective allowable feed range; screen the remaining candidate machining states in the order of machining accuracy, dynamic stability margin, allowable feed capacity, and attitude continuity of adjacent machining segments, and plan the machining state sequence for multiple consecutive machining segments; when the attitude early migration interval exists, allocate the redundant robot attitude and tool axis attitude change required after entering the stiffness drop region to the consecutive machining segments within the attitude early migration interval, so that the robot completes at least the main part of the attitude transformation before entering the stiffness drop region; Step S7: Based on the determined target machining state, compensate the theoretical five-axis toolpath. Specifically, determine the tool position compensation amount based on the normal relative deformation of multiple representative feature positions within the effective cutting area, determine the tool axis direction compensation amount based on the changing trend of the normal relative deformation between each representative feature position along the tool axis direction, and determine the target feed state within the allowable feed range to form a compensated toolpath to be verified that includes tool position, tool axis direction, and time-varying feed information. Step S8: For the toolpath to be checked and compensated, redetermine the compensated tool-workpiece contact area, cutting load, quasi-static relative deformation, and dynamic stability. If the residual contour error, dynamic stability, tool-workpiece non-interference state, and robot motion continuity do not simultaneously meet the requirements, correct the tool position, tool axis direction, or machining parameters again, and recheck. When all the check conditions are met simultaneously, determine the corresponding toolpath as the five-axis toolpath for the current machining segment. Step S9: Control the robot to complete the current machining segment according to the five-axis toolpath. During the machining process, simultaneously collect the actual cutting load, robot motion state, and vibration response that correspond to the current machining segment and tool position. After completing the predetermined machining segment, obtain at least one of the following machining results: actual machining contour, actual wall thickness, or actual material removal boundary. Based on the corresponding deviation between the actual data and the predicted data, identify the source of the deviation and selectively correct the cutting load prediction state, local dynamic response state, or material removal state model according to the source of the deviation. When the material removal state is corrected, or multiple related states simultaneously show a continuous deviation that reaches the preset reconstruction condition, redetermine the local quasi-static compliance state, local dynamic response state, tool-workpiece contact area, candidate tool axis direction, robot redundant posture, allowable feed range, and toolpath compensation amount of the subsequent machining segment, starting from the corrected actual material removal state. Step S10: Perform subsequent machining segments according to the updated machining status until the five-axis machining of the thin-walled structural component is completed.

2. The five-axis machining method for thin-walled structural parts of a robot according to claim 1, characterized in that, In step S1, when dividing the five-axis toolpath into machining segments: The tool position is obtained sequentially along the theoretical five-axis toolpath; The curvature of the machined surface, the theoretical remaining wall thickness, and the theoretical tool axis direction at each tool position are obtained respectively. When at least one of the surface curvature change, theoretical remaining wall thickness change, and theoretical tool axis direction change between adjacent tool positions meets the corresponding preset segmentation conditions, the machining segment boundary is established. The material removal state, local structural state, and tool-workpiece contact state within the same processing section are kept within a predetermined range of variation.

3. The five-axis machining method for thin-walled structural parts of a robot according to claim 1, characterized in that, In step S2: The local quasi-static compliance state includes at least the unit load displacement response of the current machining region along the normal to the machined surface and the direction of the main cutting load; The local dynamic response state includes at least the dynamic compliance characteristics of the current machining area along the normal of the machined surface, the feed direction and the tool axis direction, at least one low-order natural frequency and the corresponding damping characteristics. Multiple material removal state samples are pre-established based on different remaining wall thicknesses, local boundary constraints, and remaining material distributions, and the local quasi-static compliance state and local dynamic response state corresponding to each material removal state sample are obtained. During the formal processing, the corresponding state sample is selected based on the current material removal state, or the initial state parameters of the current processing area are obtained based on adjacent state samples, and the state parameters corresponding to the subsequent processing sections are corrected using actual processing data.

4. The five-axis machining method for thin-walled structural parts of a robot according to claim 1, characterized in that, Determining the attitude advance migration interval in step S2 includes: Based on the current actual material removal status and the predetermined processing sequence, the predicted remaining material status after the completion of multiple future processing stages is generated sequentially. The predicted local dynamic response state of the corresponding region is determined based on the predicted remaining material state of each region. Compare the local stiffness changes between adjacent future processing sections; When the stiffness decrease or the stiffness decrease rate reaches the preset sudden drop condition, the corresponding processing section will be identified as the stiffness sudden drop region. Search upstream of the stiffness drop region along the processing direction for a processing segment that can meet the requirements of continuous robot posture change and processing stability, and determine the continuous processing segment obtained by the search as the posture advance migration interval.

5. The five-axis machining method for thin-walled structural parts of a robot according to claim 1, characterized in that, Before the actual processing, multiple calibration poses are selected in the robot workspace covering the actual processing area. Known loads in multiple directions are applied to the robot end effector and the pose changes of the robot end effector are measured to obtain the equivalent stiffness parameters of the robot joints. For each candidate machining state, the robot end effector flexibility is obtained based on the current joint position of the robot and the equivalent stiffness parameters of the robot joint, and the robot end effector flexibility is converted to the tool-workpiece contact area; The equivalent compliance of the machining spindle and milling cutter is added to the converted robot end effector compliance to determine the overall compliance state of the tool side. The tool side quasi-static displacement includes at least the translation of the robot end effector, the additional displacement of the tool position caused by the rotation of the robot end effector, the elastic deformation of the machining spindle, and the bending deformation of the milling tool.

6. The five-axis machining method for thin-walled structural parts of a robot according to claim 1, characterized in that, In step S3, the tool-workpiece contact area is redefined for each candidate tool axis direction, including: Determine the boundary of the remaining material entity based on the current material removal state model; Determine the position where the cutting edge enters the remaining material, the position where it leaves the remaining material, and the effective cutting range along the tool axis in the current candidate tool axis direction; For ball end mills, the actual spherical area involved in cutting is determined based on the positional relationship between the spherical cutting edge and the remaining material entity. The corresponding cutting speed state and instantaneous material removal state are determined based on the actual position of the cutting edge involved in the cutting. The cutting load is determined separately for different candidate tool axis directions, and the same cutting load is not directly used for deformation evaluation in different candidate tool axis directions.

7. The five-axis machining method for thin-walled structural parts of a robot according to claim 1, characterized in that, The cutting balance state in step S4 is determined as follows: The cutting load obtained based on the rigid geometric contact region is used as the first predicted cutting load; The quasi-static displacement of the tool side and the quasi-static displacement of the thin-walled structural component side are determined by the first predicted cutting load. The predicted actual position of the tool relative to the thin-walled structure is changed based on the quasi-static relative deformation of the two, and the tool-workpiece contact area and predicted cutting load for the next cutting are redefined. Repeatedly perform the update of the contact area, the update of the cutting load, and the update of the quasi-static relative deformation; The iteration ends when the representative cutting load change and normal relative deformation change obtained in two consecutive calculations are both less than the corresponding convergence allowable value. If the convergence condition is not met even after reaching the preset maximum number of iterations, the candidate machining state is determined to be an unstable calculation state, and the calculation is re-performed by changing the robot's redundant posture, the tool axis direction, or reducing the cutting parameters.

8. The five-axis machining method for thin-walled structural parts of a robot according to claim 1, characterized in that, In step S5: The dynamic responses of the tool side and the thin-walled structural part side along the normal to the machined surface, the feed direction, and the tool axis direction are obtained respectively; The responses on both sides are transformed to a unified coordinate reference to obtain the relative dynamic response in the corresponding direction; When the relative vibration of the tool axis direction does not reach the dynamic contact triggering condition, the dynamic stability is evaluated by using the cutting contact area under the cutting balance state. When the relative vibration in the direction of the tool axis reaches the dynamic contact trigger condition, the axial cutting area is divided into a relatively stable contact area that continuously participates in cutting and a dynamic contact area that periodically enters or exits the material due to relative vibration, and the dynamic cutting load is re-determined based on the dynamic contact area. The dynamic stability of the corresponding candidate machining state is determined again based on the corrected dynamic cutting load.

9. The five-axis machining method for thin-walled structural parts of a robot according to claim 1, characterized in that, The allowable feed range for each candidate processing state is determined as follows: The motion state of each joint of the robot as the feed speed changes is determined based on the robot's current candidate posture and processing direction, and the first feed range under the robot's motion constraints is determined based on the robot's joint position, allowable speed and allowable acceleration. Starting from the preset minimum evaluation feed, the candidate feed rate is gradually increased within the first feed range, and the cutting load and three-dimensional dynamic response are re-predicted for each candidate feed rate; When increasing the candidate feed rate causes the predicted vibration response in any direction to exceed the corresponding allowable value, stop increasing the feed rate and determine the feed rate that still meets the vibration requirements as the allowable feed limit for the current candidate machining state. The allowable feed range is determined based on the preset minimum stable machining feed and the allowable feed upper limit.

10. The five-axis machining method for thin-walled structural parts of a robot according to claim 1, characterized in that, When screening candidate machining states in step S6, contour error, dynamic stability, no interference between tool and workpiece, and robot motion constraints are first used as hard constraints, and candidate machining states that do not meet any hard constraints are deleted. For candidate machining states that satisfy all hard constraints, the machining state with smaller relative deformation in the predicted normal direction is preferred. When the predicted normal relative deformation of multiple candidate machining states is within the same preset accuracy range, the machining state with a larger dynamic stability margin is selected first. When the dynamic stability margins of multiple candidate processing states all meet the same preset stability range, the processing state with the larger allowable feed upper limit is selected first. When multiple candidate machining states still exist, select the machining state with smaller changes in robot joints and tool axis direction compared to adjacent machining segments.

Images