A machining parameter automatic optimization method and system for a five-axis numerical control machine tool

Abstract

The present application relates to the technical field of industrial adaptive control, in particular to a machining parameter automatic optimization method and system for five-axis CNC machine tools. The specific implementation process includes: discretely analyzing machine tool parameters containing tool axis vector and translational axis position coordinates to generate time-varying geometric characteristics; constructing a CNC machine tool simulation model according to a material constitutive equation and a frequency domain response function, calculating a regenerative chatter prediction sequence, and dividing a real-time parameter feasible region in combination with global stiffness and local flexibility data; performing global iterative optimization on a machining parameter function group in the feasible region by using a multi-objective optimization algorithm to lock a Pareto optimal solution set for automatic optimization of the machining process. The present application avoids chatter and errors caused by unbalanced stiffness matching by constructing a simulation model containing physical dynamic characteristics and a real-time parameter feasible region, and simultaneously realizes the collaborative optimization of multi-objective variables, thereby significantly improving the productivity efficiency of the CNC machine tool under the premise of ensuring machining accuracy.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of industrial adaptive control, in particular to a machining parameter automatic optimization method and system for a five-axis numerical control machine tool. BACKGROUND

[0002] Five-axis linkage numerical control equipment occupies a core position in the digital manufacturing of complex curved surface parts due to its multi-degree-of-freedom trajectory interpolation and posture control capability. In the field of numerical control programming and control system, the setting of process control parameters is a key link for generating a numerical control machining program, which directly determines the efficiency of instruction execution and the final machining precision. In the prior art, for the determination of control parameters, a processing method based on an offline geometric simulation model is generally adopted. This method simulates the spatial geometric interference process of the tool when running along the predetermined tool position trajectory through computer algorithm, and calculates the instantaneous material removal rate or tool contact area at discrete time points. Based on the mapping assumption that the control load and the geometric removal volume are positively correlated, the control system will adaptively modulate the feed speed instruction in the numerical control program: reduce the instruction speed in the data segment with large geometric removal amount to meet the load constraint, and increase the instruction speed in the small removal amount or finishing data segment to shorten the cycle time, so as to try to realize program optimization control based on constant load expectation at the algorithm level.

[0003] However, the prior art has inherent defects in actual application. The complex nonlinear physical dynamic characteristics in the process control, such as time-varying fluctuation of cutting force vector, thermodynamic cumulative effect and frequency domain response characteristics of system structure, are ignored, which may cause trajectory tracking error or dynamic instability (such as chatter) in actual execution of the theoretically optimized control instruction due to stiffness mismatch. Due to the lack of accurate digital prediction of process state boundary, in order to avoid the risk of control failure, the existing strategy usually sets the global control parameters based on the conservative constraint boundary of the worst case, which causes the output power of the driving system and the dynamic performance of the feed axis to be unable to be fully utilized in most load redundant program segments, seriously limiting the production efficiency of high-end numerical control equipment. In summary, the existing control logic usually only takes a single objective function (such as time minimization or load constantization) as the convergence orientation, lacks a calculation architecture capable of globally optimizing multiple target variables such as machining efficiency, surface topography index, actuator life consumption and energy efficiency ratio which are mutually coupled and conflicting, and is difficult to meet the individualized optimal control requirements for specific task demands in high-precision manufacturing scenarios.

[0004] Therefore, a machining parameter automatic optimization method and system for a five-axis numerical control machine tool are proposed. SUMMARY

[0005] The application aims to provide a machining parameter automatic optimization method and system for a five-axis numerical control machine tool.

[0006] To achieve the above-mentioned purpose, the application provides the following technical scheme.

[0007] A machining parameter automatic optimization method for a five-axis numerical control machine tool, comprising:

[0008] Discretely analyzing machine tool parameters containing tool axis vectors and translational axis position coordinates in a machining process to generate a discrete tool position point sequence; using geometric Boolean operation to remove materials in the time domain for the discrete tool position point sequence to generate time-varying geometric features containing instantaneous contact area and material removal rate gradient;

[0009] Building a numerical control machine tool simulation model according to a material constitutive equation and a frequency domain response function, and receiving the time-varying geometric features to calculate a regenerative chatter prediction sequence representing dynamic stability of the machining process; using the regenerative chatter prediction sequence to constrain dynamic feed speed in a control parameter space according to current global stiffness data and local flexibility data of the machine tool, and dividing a real-time parameter feasible region varying with five-axis posture;

[0010] In the real-time parameter feasible region, using a multi-objective optimization algorithm to globally and iteratively optimize a machining parameter function group containing machining efficiency, surface profile error index and machine tool energy efficiency ratio; locking the machining parameter function group corresponding to a Pareto optimal solution set in the iterative optimization process to automatically optimize the machining process of the five-axis numerical control machine tool.

[0011] Preferably, the specific implementation process of discretely analyzing machine tool parameters containing tool axis vectors and translational axis position coordinates in a machining process to generate a discrete tool position point sequence comprises:

[0012] Receiving machine tool parameters in a numerical control machining process, extracting an original instruction stream composed of translational axis coordinate increments and rotary axis angle values; building a forward kinematics transformation matrix based on the topological structure of a five-axis machine tool to map and convert each axis motion data in the original instruction stream from a machine tool coordinate system to a workpiece coordinate system to solve tool axis vectors and translational axis position coordinates; using parameter curve interpolation to perform trajectory smoothing fitting on the tool axis vectors and translational axis position coordinates to build continuous tool center point motion trajectories and tool axis posture change surfaces; synchronously sampling the continuous tool center point motion trajectories and posture change surfaces according to equal arc length discretization criteria; calculating tangent vectors and normal vectors at each sampling point, and time-aligning and packaging spatial position data and posture data to generate a discrete tool position point sequence.

[0013] Preferably, the material removal in time domain of the discrete tool position sequence is generated by using geometric Boolean operation, and a specific implementation process of generating a time-varying geometric feature including instantaneous contact area and material removal rate gradient comprises:

[0014] According to the time sequence logic of the discrete tool position sequence, a swept volume envelope of the tool along a motion trajectory is constructed, a Boolean difference operation of the swept volume envelope and a current machining material is performed by using a solid construction geometric algorithm, and an intermediate geometric state of the workpiece after material removal is updated in real time; a contact area boundary of a tool cutting edge and the intermediate geometric state of the workpiece at each discrete time is extracted, and an instantaneous contact area of the tool and the workpiece is obtained by calculating a curvilinear integral; an instantaneous material removal rate is calculated based on a volume increment of a material entity removed in adjacent time steps, and a sequence of the instantaneous material removal rate is processed by time domain differentiation to extract a material removal rate gradient representing a sudden change trend of cutting force; and the instantaneous contact area and the material removal rate gradient are stored in association as a time-varying geometric feature.

[0015] Preferably, a numerical control machine tool simulation model is constructed according to a material constitutive equation and a frequency domain response function, and the time-varying geometric feature is received, and a specific implementation process of calculating a regenerative chatter prediction sequence representing dynamic stability of a machining process comprises:

[0016] According to the physical properties of the machining material, a thermodynamic state of a cutting deformation zone is analyzed to generate a material constitutive equation associated with a cutting speed and a feed amount; an instantaneous contact area in the time-varying geometric feature is combined to establish a frequency domain response function including a time-varying time lag term; according to the material constitutive equation and the frequency domain response function, a numerical control machine tool simulation model describing relative vibration of the tool and the machining material is constructed; the numerical control machine tool simulation model is subjected to stability discrimination and solving to generate a stability blade boundary representing a limit cutting depth; the time-varying geometric feature is mapped into a space defined by the stability blade boundary, a frequency band and a position exceeding a stability domain are identified, and a regenerative chatter prediction sequence is generated.

[0017] Preferably, according to current global stiffness data and local compliance data of the machine tool, a dynamic feed speed constraint is performed in a control parameter space by using the regenerative chatter prediction sequence, and a real-time parameter feasible region varying with a five-axis attitude is divided, and a specific implementation process comprises:

[0018] The global stiffness data is called based on the topology structure of the five-axis machine tool workspace, and the instantaneous stiffness matrix under the current machining posture is calculated by a multi-dimensional space interpolation algorithm according to the position coordinates of the translational axis in the discrete tool position sequence; the local flexibility data of the tool and workpiece contact interface is received, and the instantaneous stiffness matrix is corrected in mode coupling in combination with the regenerative chatter prediction sequence, and the dynamic feed speed constraint for suppressing regenerative chatter is generated in the control parameter space composed of spindle speed, feed speed and acceleration; the continuous parameter set satisfying the dynamic feed speed constraint is envelope fitted to divide the real-time parameter feasible region changing with the five-axis posture.

[0019] Preferably, in the real-time parameter feasible region, the specific implementation process of globally iteratively optimizing the machining parameter function group including machining efficiency, surface profile error index and machine tool energy efficiency ratio by using a multi-objective optimization algorithm includes:

[0020] The machining efficiency is quantified as the material removal volume per unit time, the surface profile error is defined as the weighted superposition of the tool path approximation error and the servo dynamic following error, and the machine tool energy efficiency ratio is represented as the energy conversion efficiency of the effective cutting power and the total input power; in the real-time parameter feasible region, a random sampling strategy is used to generate the machining parameter function group; the individuals of the machining parameter function group are hierarchically sorted according to the Pareto dominance relationship, and the offspring population is generated by selecting crossover and mutation operators; whether the individuals of the offspring population cross the boundary of the real-time parameter feasible region is monitored in real time, and the machining parameter function group is globally iteratively optimized.

[0021] Preferably, the specific implementation process of automatically optimizing the machining process of the five-axis numerical control machine tool by locking the machining parameter function group corresponding to the Pareto optimal solution set in the iterative optimization process includes:

[0022] The Pareto front solution set of the global iterative optimization output is extracted, the non-dominated solutions in the Pareto front solution set are evaluated in terms of utility in combination with the process preferences and production constraints of the current machining material, and the machining parameter function group with the highest utility value is selected; the original numerical control machining program file is read, the original instruction stream is dynamically reconstructed and updated by using the machining parameter function group, and the target numerical control code containing the optimized machining parameters is generated; the target numerical control code is converted into servo drive instructions by a machine tool control interpreter, and the servo motors of each axis are controlled to continue machining after adaptive parameter adjustment.

[0023] A machining parameter automatic optimization system for a five-axis numerical control machine tool includes:

[0024] A parameter processing module discretely analyzes machine tool parameters containing tool axis vectors and position coordinates of translational axes in a machining process to generate a discrete tool position point sequence; and a material removal is performed on the discrete tool position point sequence in a time domain by using a geometric Boolean operation to generate time-varying geometric features containing instantaneous contact areas and material removal rate gradients;

[0025] A machine tool control module constructs a numerical control machine tool simulation model according to a material constitutive equation and a frequency domain response function, receives the time-varying geometric features, and calculates a regenerative chatter prediction sequence representing dynamic stability of the machining process; and performs dynamic feed speed constraints in a control parameter space by using the regenerative chatter prediction sequence according to current global stiffness data and local flexibility data of the machine tool, and divides a real-time parameter feasible region varying with a five-axis posture.

[0026] An automatic optimization module performs global iterative optimization on a machining parameter function group containing machining efficiency, surface profile error indicators, and machine tool energy efficiency ratios in the real-time parameter feasible region by using a multi-objective optimization algorithm; locks the machining parameter function group corresponding to a Pareto optimal solution set in the iterative optimization process, and automatically optimizes the machining process of the five-axis numerical control machine tool.

[0027] Compared with the prior art, the present application has the following beneficial effects:

[0028] 1. The present application considers complex nonlinear physical dynamic characteristics (such as time-varying fluctuations of cutting force, thermodynamic accumulation, and frequency domain response) in the machining process, constructs a numerical control machine tool simulation model, and generates a regenerative chatter prediction sequence to realize accurate prediction of the process state boundary. This avoids trajectory tracking errors or dynamic instability (chatter) caused by imbalance of stiffness matching, and ensures machining accuracy.

[0029] 2. The present application does not use a conservative constraint boundary based on the worst case to set global control parameters, but divides a real-time parameter feasible region varying with a five-axis posture according to current global stiffness and local flexibility of the machine tool. This makes full use of the dynamic performance of the drive output power and the feed axis, improves the efficiency limited problem caused by load redundancy, and improves the productivity of the numerical control machine tool.

[0030] 3. The present application establishes a machining parameter function group containing machining efficiency, surface profile error indicators, and machine tool energy efficiency ratios. By Pareto global iterative optimization, the collaborative optimization of these mutually coupled and conflicting target variables is realized, which can meet the control requirements of machining task demand in high-precision manufacturing scenarios. BRIEF DESCRIPTION OF DRAWINGS

[0031] Fig. 1 A five-axis numerical control machine tool machining parameter automatic optimization method flowchart is proposed for the present application;

[0032] Fig. 2 A five-axis NC machine tool-oriented machining parameter automatic optimization system structure diagram is proposed for the present application;

[0033] Fig. 3 A NC machine tool machining parameter automatic optimization process diagram is proposed for the present application. DETAILED DESCRIPTION

[0034] In order to make the purpose, technical scheme and advantages of the present application more clear, the present application will be described in detail below with reference to the drawings and in combination with specific embodiments. It must be understood that the specific embodiments described herein are only used to explain the present application, and are not intended to constitute any form of limitation on the scope of protection of the present application. Therefore, all equivalent changes or modifications conceived by those of ordinary skill in the art based on the content disclosed in the present application without creative labor shall fall within the scope of protection requested by the present application.

[0035] Referring to Figs. 1 to 3 , the present application proposes a five-axis NC machine tool-oriented machining parameter automatic optimization method and system, and the technical scheme is as follows:

[0036] Embodiment one:

[0037] Referring to Fig. 1 , the present embodiment proposes a five-axis NC machine tool-oriented machining parameter automatic optimization method, which comprises:

[0038] Discretely analyzing the machine tool parameters containing the tool axis vector and the translational axis position coordinates in the machining process to generate a discrete tool position point sequence; using geometric Boolean operation to remove the material in the time domain for the discrete tool position point sequence to generate time-varying geometric characteristics containing instantaneous contact area and material removal rate gradient;

[0039] According to the material constitutive equation and the frequency domain response function, a NC machine tool simulation model is constructed, and the time-varying geometric characteristics are received to calculate a regenerative chatter prediction sequence representing the dynamic stability of the machining process; according to the current global stiffness data and local flexibility data of the machine tool, dynamic feed speed constraints are performed in the control parameter space using the regenerative chatter prediction sequence, and a real-time parameter feasible region varying with the five-axis attitude is divided;

[0040] In the real-time parameter feasible region, a multi-objective optimization algorithm is used to perform global iterative optimization on a machining parameter function group containing machining efficiency, surface contour error index and machine tool energy efficiency ratio; locking the machining parameter function group corresponding to the Pareto optimal solution set in the iterative optimization process, the machining process of the five-axis NC machine tool is automatically optimized.

[0041] Further, the machine tool parameters including the tool axis vector and the translational axis position coordinates in the machining process are discretely analyzed, and the specific implementation process of generating the discrete tool position point sequence includes:

[0042] The machine tool parameters in the numerical control machining process are received, and an original instruction stream composed of translational axis coordinate increments and rotary axis angle values is extracted. A forward kinematics transformation matrix is established based on the topology structure of the five-axis machine tool, each axis movement data in the original instruction stream is mapped and converted from the machine tool coordinate system to the workpiece coordinate system, and the tool axis vector and the translational axis position coordinates are solved. The tool axis vector and the translational axis position coordinates are fitted by using parameter curve interpolation for trajectory smoothing, and a continuous tool center point movement trajectory and a tool axis posture change surface are constructed. According to the equal arc length discretization criterion, the continuous tool center point movement trajectory and the posture change surface are synchronously sampled. Tangential vectors and normal vectors at each sampling point are calculated, and spatial position data and posture data are time-aligned and encapsulated to generate a discrete tool position point sequence.

[0043] Specifically, real-time machine tool parameters in the numerical control machining process are received, and an original instruction stream composed of X, Y, and Z three linear moving axis translational axis coordinate increments and A and C two rotary axis rotary axis angle values is extracted from the machine tool parameters. In order to convert these axis instructions relying on the physical structure of the machine tool into spatial data directly related to the geometry of the workpiece, a forward kinematics transformation matrix is established based on the topology structure of the five-axis machine tool. The forward kinematics transformation matrix converts each axis movement data in the original instruction stream from the machine tool coordinate system to the workpiece coordinate system through matrix multiplication operation containing rotation and translation factors. For the AC cradle type five-axis machine tool, this operation can eliminate the influence of the offset distance between the rotary axis center and the workpiece origin, and the A axis and C axis data originally represented as angle values, together with the linear axis coordinates, are solved to obtain the tool axis vector (i.e. the direction vector of the tool axis) and the translational axis position coordinates (i.e. the spatial coordinates of the tool tip point or the tool center point) in the workpiece coordinate system. If the original instruction stream shows that the A axis rotates 30 degrees and the C axis rotates 45 degrees, after the transformation matrix processing, it will be converted into a unit direction vector composed of three components and the corresponding three-dimensional spatial coordinate point.

[0044] Since the original instruction stream is limited by the interpolation period of the numerical control system, the point set directly converted often presents a broken line-like discrete feature on a micro scale, which is difficult to meet the demand of high-precision simulation. Therefore, the trajectory of the tool axis vector and the position coordinates of the translational axis is smoothed and fitted by using parameter curve interpolation, and the non-uniform rational B-spline algorithm is adopted to construct a continuous tool center point motion trajectory with second-order geometric continuity by taking the discrete coordinate points as control points, and the end points of the tool axis vector are interpolated on the unit sphere to construct a continuous tool axis posture change surface. This process effectively fills the gap between the original instruction points and eliminates the small position mutations. According to the equal arc length discretization criterion, the continuous tool center point motion trajectory and the posture change surface are sampled synchronously. The equal arc length discretization criterion sets a small step threshold (for example, the sampling step is set to 0.05 mm), and no matter how fast or slow the tool actually moves, the sampling points are strictly taken on the fitted curve according to the spatial distance. This strategy ensures that enough dense sampling points can be obtained at the corner of a complex curved surface with sharp curvature change, and avoids data redundancy in flat areas. After obtaining the sampling points, the tangential vector and normal vector at each sampling point are further calculated, and the spatial position data, posture data and time stamp data calculated by the feed speed are time-aligned and packaged to generate a discrete tool position point sequence.

[0045] In this embodiment, the axis motion instruction depending on a specific machine tool structure is converted into a general high-precision spatial geometric trajectory through forward kinematics transformation and parameter curve fitting, the problem of geometric feature loss caused by sparse or uneven sampling points in discrete simulation is improved, reliable geometric boundary input is provided for calculating the instantaneous material removal rate and cutting force fluctuation, and the accuracy of the whole adaptive optimization is significantly improved.

[0046] Further, the material removal of the discrete tool position point sequence in the time domain is carried out by using geometric Boolean operation, and the specific implementation process of generating time-varying geometric features including instantaneous contact area and material removal rate gradient includes:

[0047] According to the time sequence logic of the discrete tool position point sequence, a swept volume envelope of the tool along the motion trajectory is constructed, a Boolean difference operation between the swept volume envelope and the current machining material is performed by using a solid construction geometry algorithm, and the intermediate geometric state of the workpiece after material removal is updated in real time; the contact area boundary of the tool cutting edge and the intermediate geometric state of the workpiece at each discrete time is extracted, and the instantaneous contact area of the tool and the workpiece is obtained by calculating the curved surface integral; the instantaneous material removal rate is calculated based on the volume increment of the removed material entity within the adjacent time step, and the sequence of the instantaneous material removal rate is processed by time domain differentiation to extract the material removal rate gradient representing the cutting force mutation trend; and the instantaneous contact area and the material removal rate gradient are stored in association as time-varying geometric features.

[0048] Specifically, data from two adjacent time steps in the discrete tool position sequence are read. Based on the actual geometric parameters of the tool (e.g., a ball end mill model with a diameter of 12 mm), the volume of the solid swept by the tool within this small time interval is reconstructed in three-dimensional space. Since the tool position sequence has undergone high-precision interpolation processing, the swept volume envelope can accurately represent the spatial occupancy of the tool under complex actions such as cornering and helical cutting. A solid construction geometry algorithm is used to perform Boolean difference operations between the swept volume envelope and the current machining material. During this process, a dynamic solid model representing the current state of the workpiece is maintained, with the initial state being the blank model. For each time step, the generated swept volume envelope is used as the subtraction object and subtracted from the workpiece model, thereby updating the intermediate geometric state of the workpiece after material removal in real time. This step not only simulates the process of material removal but also retains the residual peak height and sidewall morphology left after cutting at the previous moment. While completing the Boolean operation, the contact area boundary between the tool cutting edge and the intermediate geometric state of the workpiece at each discrete moment is extracted, and the closed curve loop intersecting the tool solid surface and the workpiece solid surface is identified through topological traversal. The surface integral of the tool surface inside the closed curve loop is calculated to obtain the instantaneous contact area between the tool and the workpiece. In a preferred embodiment, when the tool moves from straight cutting to a circular arc, the calculated instantaneous contact area increases dramatically from 15 square millimeters to 45 square millimeters within 0.05 seconds. This precise value obtained through geometric analysis can directly reflect the drastic change in tool load.

[0049] The instantaneous material removal rate is calculated based on the volume increment of the removed material entities within adjacent time steps. The volume difference of the workpiece entity before and after the Boolean difference operation is calculated, and this difference is the chip volume generated within the current time step. Dividing this by the time step (e.g., 1 millisecond) yields the volume removal rate in cubic millimeters per second. To capture transient impacts during machining, the sequence of instantaneous material removal rates is subjected to time-domain differentiation, i.e., the rate of change of the removal rate over time is calculated, thereby extracting the material removal rate gradient characterizing the abrupt change in cutting force. When the tool cuts into hard inclusions or encounters a sudden change in cutting depth, although the instantaneous removal rate has not yet reached its peak, the material removal rate gradient will first show a high-amplitude pulse signal (e.g., the gradient value instantly reaches 5000 cubic millimeters per square second), which provides a good early warning indicator. The instantaneous contact area calculated above is correlated with the material removal rate gradient according to a unified time reference and stored as a time-varying geometric feature.

[0050] The data structure of the time-varying geometric feature is defined as a synchronous time sequence matrix or structure array containing multi-dimensional information, each row of data of which corresponds to a discretized time step, and the row index strictly follows the time sequence of the machining process. Each column of the data structure respectively stores time-aligned key variables, including absolute time stamps converted based on the current feed speed and equal-arc-length sampling step, three-dimensional coordinate values of the tool center point in the workpiece coordinate system, direction cosine components of the tool axis vector, calculated instantaneous contact area scalar values, and material removal rate gradient scalar values. The sampling frequency is dynamically determined by the aforementioned equal-arc-length discretization criterion (such as 0.05 mm) and the real-time instruction feed speed, and the time resolution is automatically increased in areas with low feed speed or large geometric curvature. The time interval between data points is adaptively adjusted according to the change in speed. In terms of storage format, a double-precision floating-point data table or a binary stream file is used for packaging, and by traversing the time index of the sequence, the geometric load data and kinematic data at the same time can be directly read as synchronous inputs of the differential equation set.

[0051] The present embodiment can identify features containing physical mechanics from pure geometric motion instructions through Boolean operations and time-domain differential processing based on entity construction geometric algorithms. The introduction of instantaneous contact area and material removal rate gradient can more sensitively identify the common cutting-in, cutting-out impact and overcut risk in five-axis machining, thereby improving the problem of tool breakage or surface vibration caused by the lag in response to dynamic load changes in complex surface machining, and significantly improving the pertinence of subsequent process parameter optimization and the stability of the machining process.

[0052] Further, a numerical control machine tool simulation model is constructed according to a material constitutive equation and a frequency domain response function, and receives the time-varying geometric feature, and the specific implementation process of calculating a regenerative chatter prediction sequence representing the dynamic stability of the machining process includes:

[0053] According to the physical properties of the machining material, the thermodynamic state of the cutting deformation zone is analyzed to generate a material constitutive equation related to the cutting speed and the feed amount. In combination with the instantaneous contact area in the time-varying geometric feature, a frequency domain response function containing a time-varying time delay term is established. According to the material constitutive equation and the frequency domain response function, a numerical control machine tool simulation model describing the relative vibration of the tool and the machining material is constructed. The numerical control machine tool simulation model is subjected to stability discrimination and solving to generate a stability blade boundary representing the limit cutting depth. The time-varying geometric feature is mapped into the space defined by the stability blade boundary to identify the frequency bands and positions that exceed the stability domain, and a regenerative chatter prediction sequence is generated.

[0054] Specifically, the establishment of the material constitutive equation relies on the determination of Johnson-Cook model constants (including yield strength, hardening modulus, strain rate sensitivity coefficient and thermal softening index) of the workpiece material through Hopkinson pressure bar experiments; in the calculation process, an iterative numerical solution strategy is adopted, the initial flow stress is estimated according to the current cutting speed and shear strain rate, the temperature rise in the shear zone is calculated based on the principle of plastic work conversion into heat, the thermal softening effect term in the constitutive model is corrected using the updated temperature value, and the temperature and stress calculation are converged until the temperature and stress calculation are converged; the obtained dynamic flow stress values are combined with the orthogonal cutting or oblique cutting mechanics model to be used for analytical calculation of the cutting force coefficients under different cutting states, and the micro-mechanical behavior of the material under high temperature and high strain rate is converted into the macroscopic cutting force vector acting on the tool teeth. Combined with the instantaneous contact area in the time-varying geometric characteristics, a frequency domain response function containing a time-varying time delay term is established. The specific construction process is as follows: due to the continuous change of the tool axis posture in five-axis machining, the engagement position of the tool and the workpiece is floating in space, resulting in the evolution of the modal parameters (mass, damping, stiffness) with time. The instantaneous contact area calculated in the last step is used to determine the projection direction of the cutting force in the tool modal coordinate system, and a time delay factor determined by the spindle speed and the number of tool teeth is introduced to represent the regenerative feedback effect generated when the current cutting tooth cuts the corrugated surface left by the last tooth.

[0055] On this basis, the nonlinear cutting force coefficients calculated according to the material constitutive equation and the structural dynamic characteristics described by the frequency domain response function are used to construct a numerical control machine tool simulation model describing the relative vibration of the tool and the machined material. The model is essentially a closed-loop time-delay differential equation that simulates the dynamic interaction process of cutting force exciting structural vibration, and in turn, the structural vibration modulating the cutting thickness. The numerical control machine tool simulation model is solved by stability discrimination, and the semi-discrete method is used to perform eigenvalue analysis in the parameter plane of spindle speed and axial cutting depth, generating a stability blade boundary representing the limit cutting depth. The boundary curve divides the parameter space into stable and unstable regions. The time-varying geometric features are mapped into the space defined by the stability blade boundary, and the actual cutting depth and spindle speed at each discrete time step are read as coordinate points and projected onto the stability blade graph. If the coordinate point at a certain time falls above the stability boundary curve, it is identified as a chatter risk area, and the specific chatter frequency and amplitude growth rate are calculated. The frequency bands and positions that exceed the stable region are marked in chronological order to generate a regenerative chatter prediction sequence. The regenerative chatter prediction sequence is constructed as a state data table corresponding to the discrete tool position sequence, with each record containing a timestamp or path position index corresponding to the machining time, and a Boolean flag indicating the stability of the current cutting state. For records marked as unstable, the sequence further encapsulates the dominant chatter frequency calculated by eigenvalue analysis, the limit cutting depth threshold at the current speed, and the amplitude growth rate parameter representing the degree of negative damping of the system. The sequence not only records whether chatter occurs, but also quantifies how chatter occurs, i.e., it identifies the specific frequency components that cause instability and the amplitude ratio that exceeds the stability boundary, providing precise numerical basis and control interface for subsequent modules to avoid resonance frequencies by adjusting spindle speed or reduce cutting gain by reducing feed speed.

[0056] The present embodiment breaks through the limitations of static stiffness checking by combining the thermodynamic properties of the material constitutive equation and the kinetic properties of the time-varying time delay. The generation of the regenerative chatter prediction sequence enables accurate prediction of self-excited vibration caused by process system dynamic characteristics at the physical level before the execution of numerical control code, thereby improving the frequent tool damage caused by blind pursuit of efficiency in complex surface machining, and maximizing material removal rate under the premise of ensuring surface quality.

[0057] Further, according to the current global stiffness data and local flexibility data of the machine tool, dynamic feed speed constraints are performed in the control parameter space using the regenerative chatter prediction sequence, and the specific implementation process of dividing the real-time parameter feasible region that changes with the five-axis attitude includes:

[0058] The global stiffness data is called based on the topology of the five-axis machine tool workspace, and the instantaneous stiffness matrix under the current machining posture is calculated by a multi-dimensional space interpolation algorithm according to the translational axis position coordinates in the discrete tool position sequence; the local flexibility data of the tool and workpiece contact interface is received and combined with the regenerative chatter prediction sequence to correct the modal coupling of the instantaneous stiffness matrix, and a dynamic feed speed constraint for suppressing regenerative chatter is generated in the control parameter space composed of spindle speed, feed speed and acceleration; the continuous parameter set satisfying the dynamic feed speed constraint is envelope fitted to divide the real-time parameter feasible region varying with the five-axis posture.

[0059] Specifically, the static stiffness field data of the machine tool at different axial positions (such as ram extension length, beam span position) is stored in advance, and this database is usually obtained by finite element analysis or on-site calibration by a laser interferometer. The translational axis position coordinates in the discrete tool position sequence are read, and the specific three-dimensional coordinates of the current tool center point in the machine tool workspace are identified (for example: 2500 mm at the X-axis stroke, 800 mm at the Z-axis ram down). By a multi-dimensional space interpolation algorithm, such as Kriging interpolation or trilinear interpolation, calculation is performed in the discrete stiffness field data to calculate the instantaneous stiffness matrix under the current machining posture. In a preferred embodiment, when the Z-axis ram is extended from 200 mm to 800 mm, the tool tip static stiffness calculated by interpolation may decrease significantly from 60 N / μm to 25 N / μm. This position-dependent stiffness change is the physical basis for subsequent optimization.

[0060] The local flexibility data of the tool-workpiece contact interface, which contains the flexibility characteristics of the tool holder and the overhang part of the tool, is received, and the modal coupling correction is performed on the instantaneous stiffness matrix in combination with the regenerative chatter prediction sequence. Since there is a series coupling relationship between the machine tool rigidity and the local flexibility of the tool, the local flexibility matrix is superimposed on the global stiffness matrix through the frequency response function synthesis, so as to obtain the total dynamic stiffness at the machining point. In the control parameter space composed of the spindle speed, the feed speed and the acceleration, the dynamic feed speed constraint for suppressing the regenerative chatter is generated. For example, when the low-rigidity region (Z-axis deep cavity machining) is detected and the chatter prediction sequence prompts that there is an unstable mode of 420 Hz, instead of relying only on the spindle speed adjustment, the maximum feed per tooth allowed to maintain the cutting force within the stable limit is calculated, and then the upper limit of the dynamic feed speed is derived. Assuming that the original feed speed is 5000 mm / min, the constraint condition after the rigidity correction may force it to be reduced to 2800 mm / min, while limiting the axial acceleration to not more than 0.2g to prevent the inertial force from inducing structural resonance. The continuous parameter set satisfying the dynamic feed speed constraint is envelope fitted to divide the real-time parameter feasible region varying with the five-axis posture. The envelope fitting process adopts the piecewise linear fitting technology to construct a closed geometric polyhedron or curved surface space in the three-dimensional Euclidean space composed of the spindle speed, the feed speed and the axial acceleration; the boundary constraint function of the feasible region is composed of two parts, one part is the physical limit constant inherent to the machine tool servo drive system, including the maximum spindle speed, the maximum combined feed speed of each axis and the maximum allowable acceleration, and the other part is the dynamic stability boundary converted based on the foregoing regenerative chatter prediction sequence, which is a nonlinear function relationship between the feed speed and the spindle speed, that is, in a specific spindle speed interval, the maximum feed speed allowed is strictly limited by the stable blade limit depth of cutting at this speed. The real-time parameter feasible region is finally defined as the intersection of a group of linear or polynomial inequality groups, and only needs to combine the randomly generated candidate parameter combination into the inequality group for logical judgment. If all the inequalities are true at the same time, it is determined that the parameter combination is located in the feasible region, otherwise it is determined as out of range and triggers the constraint punishment or boundary repair mechanism.

[0061] The embodiment improves the problem of ignoring the influence of machine tool posture change on dynamic performance in numerical control machining by introducing position-based global stiffness interpolation and modal coupling correction. The division of the real-time parameter feasible region allows the high feed performance to be fully utilized in the region with good machine tool rigidity, and strict constraints are automatically applied in the region with weak rigidity, so that the overall machining efficiency and machine tool productivity of large and complex structural parts are significantly improved on the premise of ensuring no chatter and dynamic stability throughout the machining.

[0062] Further, in the real-time parameter feasible region, the specific implementation process of global iterative optimization of the machining parameter function group containing machining efficiency, surface profile error index and machine tool energy efficiency ratio by using multi-objective optimization algorithm includes:

[0063] The machining efficiency is quantified as the material removal volume per unit time, the surface profile error is defined as the weighted superposition of the tool path approximation error and the servo dynamic following error, and the machine tool energy efficiency ratio is represented as the energy conversion efficiency of the effective cutting power and the total input power of the drive; under the real-time parameter feasible region, a random sampling strategy is used to generate a machining parameter function group; the individuals of the machining parameter function group are hierarchically sorted according to the Pareto dominance relationship, and the offspring population is generated by selecting crossover and mutation operators; the global iterative optimization of the machining parameter function group is monitored in real time whether the individual of the offspring population crosses the boundary of the real-time parameter feasible region.

[0064] Specifically, the three core objective functions involved in optimization are physically quantified and defined. The machining efficiency is quantified as the material removal volume per unit time, i.e. the product of cutting depth, cutting width and feed speed, which directly reflects the chip output capacity of the machine tool in actual calculation. The surface profile error is defined as the weighted superposition of the tool path approximation error and the servo dynamic following error, i.e. the chord height error generated by the calculation parameter curve interpolation is taken as the geometric approximation error, while the following error is predicted based on the hysteresis characteristic of the servo axis, and according to the high requirement of finishing on surface quality, the weights of the two are set to 0.3 and 0.7 respectively in this embodiment, which can be set by the person skilled in the art according to actual needs, so as to synthesize a comprehensive error evaluation index. By collecting real-time current and voltage data of the spindle and feed axis driver, the machine tool energy efficiency ratio is represented as the energy conversion efficiency of the effective cutting power (power directly used for material removal) and the total input power of the drive (total input power containing motor copper loss, iron loss and mechanical friction loss). For example: at a certain moment, the total input power is monitored to be 15 kW, while the cutting power calculated by the cutting force model is 9 kW, so the energy efficiency ratio at this moment is 60%. After the quantification of the objective function, under the real-time parameter feasible region, a random sampling strategy is used to generate a machining parameter function group. By using the Latin hypercube sampling technology, an initial population is randomly generated within the feasible range of the spindle speed (such as 8000 to 12000 rpm) and the feed speed (such as 2000 to 6000 mm / min) determined by the chatter prediction sequence, for example: containing 100 different parameter combinations. These combinations are strictly limited by the feasible region boundary, ensuring that each initial solution is physically stable and safe.

[0065] According to the Pareto dominance relationship, individual of the machining parameter function set is hierarchically sorted, any two individuals A and B in the population are compared, if individual A is not worse than individual B in the three dimensions of machining efficiency, contour error and energy efficiency ratio, and at least one dimension is better than individual B, then it is determined that A dominates B. Based on this dominance relationship, all solutions are divided into several non-dominated levels, and the solution set in the first level represents the current optimal trade-off scheme. The offspring population is generated by selecting crossover and mutation operators. The parent is selected by binary tournament selection method, the gene recombination is carried out by using simulated binary crossover operator, and the polynomial mutation is performed with a small probability (such as 0.1), so as to generate offspring individuals with new parameter characteristics. In this process, boundary constraint check is carried out, that is, whether the individual of the offspring population exceeds the boundary of the real-time parameter feasible region is monitored in real time. Since the crossover and mutation operations may cause the newly generated parameter combination (for example: very high feed speed) to jump out of the previously divided chatter-free stable region, the compliance check is performed on each offspring individual. Once it is found that the boundary is exceeded, the boundary rebound or projection repair strategy is immediately used to force the individual back to the boundary of the feasible region, or directly give a poor fitness value to eliminate it. After a predetermined number of iterations (such as 50 generations), a set of non-dominated solutions, that is, the Pareto front, is output.

[0066] The embodiment effectively improves the problem that single objective optimization (such as only pursuing the fastest speed) often leads to surface quality deterioration or energy consumption surge by combining the physical multi-dimensional constraints (efficiency, quality, energy consumption) with the mathematical multi-objective evolutionary algorithm, and strictly limiting the optimization within the dynamic stable domain. The surface error is refined into the superposition of approximation and servo error, and the energy efficiency ratio index is introduced, so that the finally optimized parameters not only process extremely fast, and avoid the motor running in the low efficiency area.

[0067] Further, the machining parameter function set corresponding to the Pareto optimal solution set in the locking iterative optimization process is locked, and the specific implementation process of automatically optimizing the machining process of the five-axis numerical control machine tool includes:

[0068] The Pareto front solution set output by the global iterative optimization is extracted, the non-dominated solutions in the Pareto front solution set are evaluated in combination with the process preference and production constraint of the current machining material, and the machining parameter function set with the highest utility value is selected. The original numerical control machining program file is read, the original instruction stream is dynamically reconstructed and updated by using the machining parameter function set, and the target numerical control code containing the optimized machining parameters is generated. The target numerical control code is converted into servo drive instructions by the machine tool control interpreter, and the servo motors of each axis are controlled to continue machining after adaptive parameter adjustment.

[0069] Specifically, a Pareto front solution set of the global iterative optimization output is extracted, which usually contains tens of sets of mathematically non-dominated parameter combinations, for example: some combinations focus on extreme machining efficiency, while some combinations focus on optimal surface quality. An evaluation model based on a weighted utility function is established in combination with the process preferences and production constraints of the current machining material. The evaluation model specifically adopts a linear weighted summation model based on normalization processing. The numerical values of each target dimension in the Pareto front solution set are subjected to maximum and minimum normalization processing, and the machining efficiency, contour error and energy efficiency ratio indicators of different dimensions are mapped to a unified dimensionless interval to eliminate the biasing influence of the order of magnitude difference on the evaluation results. According to the production scene mode, the corresponding weight distribution rule is called, for example: in the fine machining mode, the weight coefficient of the surface contour error indicator is set to a dominant value higher than 0.5, while in the rough machining mode, the weight proportion of the machining efficiency indicator is significantly increased, or in the energy saving mode, the machine tool energy efficiency ratio indicator is focused on. The sum of the product of the normalized target values and the corresponding weight coefficients is calculated as the comprehensive utility score, and the unique optimal machining parameter combination is automatically locked from the non-dominated solution set by comparing the scores, thereby converting the complex mathematical multi-objective trade-off into a single decision output that meets the actual process requirements. In a preferred embodiment, since it is a fine machining process of the mold, the surface roughness is the primary constraint, and in this embodiment, the weight coefficient of the surface contour error indicator is therefore set to 0.6, the machining efficiency weight is 0.3, and the machine tool energy efficiency ratio weight is 0.1. Those skilled in the art can set it according to actual needs. The non-dominated solutions in the Pareto front solution set are evaluated for utility, the comprehensive score of each solution is calculated, and the machining parameter function set with the highest utility value is selected from among them.

[0070] The original numerical control machining program file stored in the buffer of the numerical control system is read, and the geometric path instructions in the original code are not changed, but the original instruction stream is dynamically reconstructed and updated by using the machining parameter function group. The original code is parsed line by line, the feed speed instruction and the spindle speed instruction are identified, and the dynamic parameter values calculated by the optimization algorithm are replaced. For example, the feed speed of a certain straight line in the original code is fixed at 1500 mm / min, after reconstruction, the instruction is subdivided into multiple segments, the starting stage is set to 2000 mm / min according to the acceleration and deceleration capability, the middle stable zone is increased to 3500 mm / min, and the end of the terminal deceleration zone is automatically adjusted to 1200 mm / min. Through this fine replacement, the target numerical control code containing the optimized machining parameters is generated. The target numerical control code is sent into the numerical control kernel in real time, and the numerical control kernel converts the target numerical control code into servo drive instructions through the machine tool control interpreter. The interpreter interpolation operation is the position increment of each axis, and the corresponding current and speed control signals are generated to control the servo motor of each axis to continue processing after adaptive parameter adjustment. In actual machining verification, for a large bumper mold, the overall cutting time is shortened by 28% compared with the original fixed parameter machining, and the manual polishing time of the mold surface is reduced by about 40% due to the automatic reduction of dynamic impact at the corner.

[0071] The embodiment improves the problem that the optimization result often stays at the simulation level and is difficult to directly guide production by establishing an automatic decision-making mechanism based on the utility function and a code dynamic reconstruction technology, and realizes the connection from multi-objective optimization to servo execution. Without human intervention, the optimal control strategy can be flexibly switched according to different production tasks (such as focusing on efficiency for rough machining and focusing on quality for finish machining), which significantly improves the intelligent level and actual machining efficiency of five-axis numerical control equipment.

[0072] Embodiment two:

[0073] The embodiment provides a machining parameter automatic optimization system for a five-axis numerical control machine tool, which is applied to the five-axis numerical control machine tool, and refers to Fig. 2 The system includes a parameter processing module, a machine tool control module, and an automatic optimization module.

[0074] The parameter processing module discretely analyzes the machine tool parameters containing the tool axis vector and the position coordinates of the translational axis in the machining process, generates a discrete tool position point sequence, and uses geometric Boolean operation to remove materials in the time domain of the discrete tool position point sequence, to generate time-varying geometric characteristics containing instantaneous contact area and material removal rate gradient.

[0075] The machine tool control module constructs a simulation model of the numerical control machine tool according to a material constitutive equation and a frequency domain response function, receives the time-varying geometric characteristics, and calculates a regenerative chatter prediction sequence representing dynamic stability of a machining process; according to current global stiffness data and local flexibility data of the machine tool, dynamic feed speed constraints are performed in a control parameter space by using the regenerative chatter prediction sequence, and a real-time parameter feasible region varying with a five-axis posture is divided;

[0076] The automatic optimization module performs global iterative optimization on a machining parameter function group including machining efficiency, surface profile error indicators and machine tool energy efficiency ratio in the real-time parameter feasible region by using a multi-objective optimization algorithm; locking the machining parameter function group corresponding to a Pareto optimal solution set in the iterative optimization process, the machining process of the five-axis numerical control machine tool is automatically optimized.

[0077] Further, the parameter processing module receives or reads the numerical control machining program file in real time through the high-speed bus interface or offline. The original instruction stream composed of the prismatic axis coordinate increment and the rotary axis angle value is extracted, and the five-axis machine tool topology structure parameters (including the geometric dimensions of each axis, the rotary displacement, and the rotary center coordinates) pre-stored in the system memory are called to establish the forward kinematics transformation matrix. Through the matrix operation, the parameter processing module maps the axis motion data in the machine tool coordinate system to the workpiece coordinate system point by point, accurately calculates the three-dimensional coordinates of the tool center point and the direction cosine of the tool axis vector at each moment, and thus completes the spatial conversion from "machine tool motion" to "cutting motion". In order to overcome the discretization error caused by the interpolation period in the original instruction, the parameter processing module uses parameter curve interpolation technology, such as the non-uniform rational B-spline algorithm, to perform high-order trajectory smoothing fitting on the calculated tool axis vector and prismatic axis position coordinates, and constructs a geometrically continuous and derivable tool center point motion trajectory and tool axis attitude change surface. The module executes the equal arc length discretization criterion, sets a small sampling step (for example, 0.05 millimeters), and performs synchronous sampling on the fitted curve. This processing ensures that even at long straight line segments with sparse instructions or sharp turns with dense instructions, a uniformly dense discrete tool position point sequence can be obtained, providing a standardized data basis for subsequent contact analysis. After generating the discrete tool position point sequence, the parameter processing module uses geometric Boolean operations to simulate material removal in the time domain. The module constructs the swept volume envelope of the tool along the motion trajectory in the virtual environment, and performs Boolean difference operation between the swept volume envelope and the current machining material entity by using the entity construction geometry algorithm or the multi-level depth pixel model. This operation process updates and records the intermediate geometric state of the workpiece after the material is removed in real time. Based on this dynamic geometric model, the parameter processing module extracts the contact area boundary between the tool cutting edge and the intermediate geometric state of the workpiece at each discrete time, calculates the instantaneous contact area to the square millimeter level through a complex curved surface integral algorithm. At the same time, the module calculates the volume increment of the material entity removed in the adjacent time step, calculates the instantaneous material removal rate, and further performs time domain differential processing on the removal rate sequence to extract the material removal rate gradient representing the cutting force mutation trend. Finally, the parameter processing module time-aligns and correlates the instantaneous contact area and the material removal rate gradient, and outputs the time-varying geometric features containing high-dimensional physical information.

[0078] Further, the machine tool control module receives the time-varying geometric features including instantaneous contact area and material removal rate gradient output by the parameter processing module. The machine tool control module has a built-in database of physical properties of specific machined materials. Based on the thermodynamic properties and rheological stress characteristics of the material, the machine tool control module analyzes the thermodynamic state of the cutting deformation zone under different cutting speeds and feed rates, and generates a nonlinear material constitutive equation associated with the cutting parameters. In actual operation, this equation can describe the thermal softening effect of the material at cutting temperatures up to 500 degrees Celsius and the hardening effect at high strain rates, thereby providing an accurate physical benchmark for dynamic calculation of cutting forces. On this basis, the machine tool control module combines the instantaneous contact area in the time-varying geometric features to establish a frequency domain response function that includes time-varying time delay terms. Due to the changing relative attitude of the tool and workpiece in five-axis machining, the modal characteristics of the system are not constant. The machine tool control module simulates the regenerative feedback mechanism that occurs when the cutting edge of the tool cuts through the corrugated surface left by the previous revolution of cutting by introducing a time delay differential equation. Based on the cutting force coefficients determined by the material constitutive equation and the structural dynamic characteristics described by the frequency domain response function, the machine tool control module constructs a numerical control machine tool simulation model that describes the relative vibration of the tool and the machined material. The machine tool control module uses a semi-discrete numerical solution method to solve the stability of the simulation model in the full frequency range, generating a stability blade boundary graph that represents the limit cutting depth. The system maps the actual cutting depth and spindle speed of the current time step into the parameter space defined by the stability blade boundary, and once it identifies that the current operating point falls into the unstable region, it generates a regenerative chatter prediction sequence that includes the specific chatter frequency and amplitude growth trend. For example, in a certain deep cavity corner machining simulation, the module successfully predicted that at a spindle speed of 12000 revolutions per minute, a regenerative chatter with a frequency of 850 Hz will be induced due to a sudden change in cutting width. The machine tool control module performs stiffness checking based on position correlation. Based on the three-dimensional topological structure of the five-axis machine tool workspace, the machine tool control module calls the pre-calibrated global stiffness data field. According to the position coordinates of the translational axes (such as the length of the ram extension and the position of the beam span) in the discrete tool position sequence, the machine tool control module calculates the instantaneous stiffness matrix under the current machine tool attitude through a multi-dimensional space interpolation algorithm. At the same time, the module receives the local flexibility data of the tool and workpiece contact interface (mainly determined by the tool holder and tool overhang), and combines the regenerative chatter prediction sequence to correct the modal coupling of the instantaneous stiffness matrix, thereby obtaining the total dynamic stiffness of the system at the machining point. In the three-dimensional control parameter space composed of spindle speed, feed rate, and acceleration, the machine tool control module uses the corrected system dynamic model to generate a dynamic feed rate constraint to suppress regenerative chatter. The module performs envelope fitting on the continuous parameter set that satisfies the dynamic feed rate constraint and does not exceed the machine tool drive capacity, dividing the real-time parameter feasible region that changes with the five-axis attitude.In practical applications, the feasible region appears as a "safety pipeline" that dynamically changes with the tool path. For example, in the full extension state of the slide of a machine tool with weak rigidity, the feasible region is automatically narrowed to limit the maximum acceleration to not more than 0.15g, while in the low-position machining state with good rigidity, it is relaxed to 0.3g, thereby providing a precise search space for the subsequent automatic optimization module.

[0079] Further, the automatic optimization module quantifies the machining efficiency as the material removal volume per unit time, i.e. the real-time product of the cutting depth, cutting width and feed speed; defines the surface profile error as the weighted superposition of the tool path approximation error and servo dynamic following error, where the approximation error is derived from the chord height deviation of curve interpolation, and the following error is derived from the hysteresis response of servo system, and the weights of the two errors are dynamically assigned according to the machining accuracy level; represents the machine tool energy efficiency ratio as the energy conversion efficiency of the effective cutting power and the total input power of the drive, aiming to improve the proportion of electric energy converted into chip removal work. After the optimization objectives are clear, the automatic optimization module starts the optimization program within the real-time parameter feasible region transmitted by the machine tool control module. This module uses an improved non-dominated sorting genetic algorithm, which first generates an initial population within the spindle speed and feed speed range defined by the feasible region using the Latin hypercube sampling strategy. For example, in the finishing machining of a stainless steel impeller blade, the feasible region limits the speed to 4000-8000 rpm and the feed to 1000-3000 mm / min, and the module randomly generates 50 initial parameter schemes within this range. Subsequently, the automatic optimization module sorts the individuals of the machining parameter function set according to the Pareto dominance relationship, and classifies the solutions that cannot be simultaneously dominated by other solutions in all objectives into the first non-dominated layer. The module generates a child population by selection, simulated binary crossover and polynomial mutation operators, and in this process, the module monitors in real time whether the newly generated child individuals exceed the boundaries of the real-time parameter feasible region. Once it is detected that a certain parameter combination (such as high feed causing chatter risk) exceeds the boundaries, it immediately uses boundary mapping to pull it back within the feasible region or gives it a very low fitness value to eliminate it, ensuring that all solutions are physically stable. After a preset number of iterations converge, the automatic optimization module locks the machining parameter function set corresponding to the Pareto optimal solution set in the iterative optimization process. To achieve automatic control, the module evaluates the utility function of the non-dominated solutions in the Pareto front solution set according to the process preferences of the current machining task (such as finishing machining prioritizing surface quality), and selects the parameter set with the highest comprehensive utility value. Then, the module reads the original instruction stream in the buffer of the numerical control system, and uses the selected optimal machining parameter function set to dynamically reconstruct and update the original instructions. Specifically, the module does not change the original geometric trajectory coordinates, but modifies the feed and speed command values in real time to generate target numerical control code containing optimized machining parameters. The automatic optimization module converts the target numerical control code into bottom servo drive instructions through the machine tool control interpreter, and controls the servo motors of each axis to continue machining after adaptive parameter adjustment. In actual machining verification, for the flow finishing machining of a certain type of titanium alloy blisk, the automatic optimization module adjusts the originally constant feed speed to a variable feed speed that fluctuates with the curvature and cutting force.The measured data shows that the machining efficiency is improved by 22% under the premise of ensuring that the surface roughness is stable below 0.8 microns, and the overall energy efficiency ratio of the machine tool is improved by 15% due to the avoidance of low-efficiency area operation.

[0080] The parameter processing module in this embodiment successfully converts static numerical control code into dynamic tool-workpiece geometric interaction characteristics through high-precision kinematic reconstruction and entity Boolean operation. The time-varying geometric characteristics generated by this module not only quantify the size of the cutting load (contact area), but also acutely capture the severity of load changes (removal rate gradient), providing underlying data support for subsequent machine tool control modules for accurate chatter prediction and adaptive parameter optimization.

[0081] The machine tool control module realizes deep prediction of the physical state of the five-axis machining process by integrating material thermodynamics, regenerative chatter theory, and machine tool position-related stiffness characteristics. It can identify and avoid machining defects caused by process system dynamic characteristics (such as chatter and insufficient stiffness) in advance. By dividing the real-time parameter feasible region that changes with the attitude, this module ensures that subsequent parameter optimization is not only efficient in theory, but also stable and safe in physics, effectively solving the surface vibration and tool abnormal wear problems caused by machine tool dynamic performance fluctuations in complex surface machining.

[0082] The automatic optimization module successfully improves the long-standing coupling conflict between machining efficiency, surface quality, and energy consumption by executing a multi-objective evolutionary algorithm within physical constraints (feasible region). It can automatically find the best balance point for each performance indicator based on the real-time state of the cutting process. Through dynamic reconstruction at the code level and servo execution, it realizes a closed-loop automation of "perception-decision-control", which not only improves the machining capacity of high-end numerical control equipment, but also significantly reduces waste and energy consumption.

[0083] Example Three:

[0084] This embodiment deploys the above-mentioned machining parameter automatic optimization method and system for five-axis numerical control machine tools in the X numerical control machine tool machining workshop of a factory, and realizes automatic optimization of machining parameters by referring to Fig. 3

[0085] ​Further, the system reads the original NC program file through the high-speed data interface, extracts the original instruction stream containing the X, Y, Z linear axis coordinate increments and the A, C rotary axis angle values. Since the workshop uses a double-turntable five-axis linkage machining center, the system establishes the forward kinematics transformation matrix based on the geometric dimensions, rotation center offset and other topological structure parameters of each axis of the machine tool. Through matrix operation, the system accurately maps and converts the motion data of each axis in the original instruction stream from the machine tool coordinate system to the workpiece coordinate system, and calculates the tool axis vector and the position coordinates of the translational axis changing with time. In order to eliminate the microscopic broken line effect caused by the discretization of the original instruction, the system uses the non-uniform rational B-spline parameter curve interpolation technology to smooth and fit the tool axis vector and the position coordinates of the translational axis with high-order trajectory, and constructs the tool center point motion trajectory and the tool axis attitude change surface with second-order geometric continuity. The system executes the equal arc length discretization criterion, sets the sampling step length to 0.1 millimeter, and synchronously samples the fitted continuous trajectory and surface. The system calculates the tangent vector and normal vector at each sampling point, and time-aligns and encapsulates the calculated spatial position data and attitude data, thereby generating a discrete tool position point sequence containing about three million data points.

[0086] Further, the system uses geometric Boolean operation to remove material in the time domain based on the discrete tool position point sequence, generates time-varying geometric features including instantaneous contact area and material removal rate gradient. According to the time logic of the discrete tool position point sequence, the swept volume envelope of the 10 millimeter diameter ball end mill along the motion trajectory is constructed. Using the solid construction geometry algorithm, the system performs Boolean difference operation between the swept volume envelope and the current workpiece blank model, and updates and records the intermediate geometric state of the workpiece after material removal in real time. The system extracts the contact area boundary between the tool cutting edge and the intermediate geometric state of the workpiece at each discrete time, and calculates the instantaneous contact area between the tool and the workpiece through the surface integral algorithm. For example, during the blade root corner cleaning machining stage, the system calculates that the instantaneous contact area increases from 5 square millimeters to 18 square millimeters within 0.02 seconds. At the same time, the instantaneous material removal rate is calculated based on the volume increment of the removed material entity within the adjacent time step, and the removal rate sequence is time-domain differentiated to extract the material removal rate gradient representing the cutting force mutation trend. The system stores the instantaneous contact area and the material removal rate gradient in association, forming time-varying geometric features reflecting the geometric load changes in the machining process.

[0087] Further, the system analyzes the thermodynamic state of the cutting deformation zone under high temperature and high strain rate according to the physical properties of the machining material, and generates the Johnson-Cook material constitutive equation associated with the cutting speed and the feed rate. In combination with the instantaneous contact area, the system establishes a frequency domain response function containing time-varying time delay items, which accurately describes the regenerative feedback mechanism between the tool and the workpiece caused by relative vibration. Based on this, a numerical control machine tool simulation model describing the relative vibration between the tool and the machining material is constructed, and a semi-discrete method is used to solve the stability discrimination of the model in the full frequency band, generating a stability blade boundary representing the limit cutting depth. The system maps the time-varying geometric features into the parameter space defined by the stability blade boundary, and identifies the frequency band and position that exceed the stability domain. In actual operation, the system successfully predicts that at the outlet of the blade disc flow passage, if the original rotating speed is maintained, a regenerative chatter with a frequency of 650 Hz will occur, and the risk area is marked to generate a regenerative chatter prediction sequence.

[0088] Further, the system uses the regenerative chatter prediction sequence to perform dynamic feed speed constraint in the control parameter space according to the current global stiffness data and local flexibility data of the machine tool, and divides a real-time parameter feasible region that changes with the five-axis attitude. The system calls the pre-calibrated global stiffness data field based on the topological structure of the five-axis machine tool workspace, and calculates the instantaneous stiffness matrix under the current machining attitude through a trilinear interpolation algorithm according to the position coordinates of the translational axes in the discrete tool position point sequence. The system receives the local flexibility data of the tool system and combines the regenerative chatter prediction sequence to perform modal coupling correction on the instantaneous stiffness matrix. In the three-dimensional control parameter space composed of spindle speed, feed speed and acceleration, the system generates a dynamic feed speed constraint to suppress regenerative chatter. For example, in the full extension working condition of the weaker ram, the system limits the maximum allowable acceleration to within 0.1g. The system performs envelope fitting on the continuous parameter set that satisfies these constraint conditions to divide a real-time parameter feasible region that changes dynamically with the five-axis attitude.

[0089] Further, the system uses a multi-objective optimization algorithm to perform global iterative optimization on a machining parameter function group containing machining efficiency, surface profile error index and machine tool energy efficiency ratio in the real-time parameter feasible region. The machining efficiency is quantified as the material removal volume per unit time, the surface profile error is defined as the weighted superposition of the geometric approximation error and the servo following error, and the machine tool energy efficiency ratio is represented as the ratio of the effective cutting power to the total input power. In the real-time parameter feasible region, the system generates an initial machining parameter function group using a random sampling strategy, and sorts the individuals according to the Pareto dominance relationship. The system generates a child population by selection, crossover and mutation operators, and monitors in real time whether the child individuals exceed the boundary of the feasible region, and punishes or repairs the individuals that exceed the boundary. After 50 generations of genetic iteration, the system outputs a set of non-dominated Pareto front solutions.

[0090] Further, the system locks the machining parameter function set corresponding to the Pareto optimal solution set in the iteration optimization process, and automatically optimizes the machining process of the five-axis NC machine tool. The Pareto front solution set is extracted, and the non-dominated solution is evaluated in combination with the process preference of the high surface quality requirement of finishing, to select a group of machining parameter function set with the highest utility value. The system reads the original NC machining program file, and dynamically reconstructs and updates the original instruction stream using the preferred machining parameter function set, to generate target NC code containing variable feed speed and variable spindle speed. Through the machine tool control interpreter, the target code is converted into servo drive instructions to control the servo motors of each axis to continue machining after adaptive parameter adjustment.

[0091] The embodiment can predict regenerative chatter by constructing a physical simulation model containing material thermodynamics and time-varying dynamics, and ensure machining accuracy. The stiffness field related to the machine tool posture is used to divide the real-time parameter feasible region, to fully utilize the machine tool performance under the premise of ensuring stability, and significantly improve the machining efficiency. The multi-objective Pareto optimization algorithm is adopted to realize the collaborative optimization of machining efficiency, surface profile error and machine energy efficiency ratio, and improve the performance conflict caused by single objective optimization. The material removal rate gradient is extracted to capture transient cutting impact, and the instruction is adjusted in advance to reduce mechanical load, effectively prolonging the tool and equipment life.

[0092] It should be noted that the above-described embodiments are only exemplary, and the purpose is to help understand the present application, rather than to limit the present application. Those skilled in the art can make various changes and improvements after understanding the core idea of the present application. Therefore, the protection scope of the present application is defined by the appended claims and their equivalent principles.

Claims

1. A method for automatic optimization of machining parameters for five-axis CNC machine tools, characterized in that, include: Discretize and analyze the machine tool parameters, including the tool axis vector and translation axis position coordinates, during the machining process to generate a discrete tool position sequence; use geometric Boolean operations to remove material in the time domain from the discrete tool position sequence to generate time-varying geometric features containing instantaneous contact area and material removal rate gradient; A CNC machine tool simulation model is constructed based on the material constitutive equation and frequency domain response function. The time-varying geometric features are received, and a regenerative chatter prediction sequence characterizing the dynamic stability of the machining process is calculated. Based on the current global stiffness data and local flexibility data of the machine tool, the regenerative chatter prediction sequence is used to constrain the dynamic feed speed in the control parameter space, and the real-time parameter feasible region that varies with the five-axis attitude is delineated. Within the feasible region of the real-time parameters, a multi-objective optimization algorithm is used to perform global iterative optimization on the set of machining parameter functions, which includes machining efficiency, surface contour error index and machine tool energy efficiency ratio; the set of machining parameter functions corresponding to the Pareto optimal solution set is locked during the iterative optimization process, and the machining process of the five-axis CNC machine tool is automatically optimized.

2. The automatic optimization method for machining parameters of a five-axis CNC machine tool according to claim 1, characterized in that, The specific implementation process of discretizing and analyzing machine tool parameters, including tool axis vector and translation axis position coordinates, during the machining process to generate a discrete tool position sequence includes: The system receives machine tool parameters from the CNC machining process and extracts the original command stream consisting of translational axis coordinate increments and rotary axis angle values. Based on the topology of the five-axis machine tool, a forward kinematic transformation matrix is ​​established to map and transform the motion data of each axis in the original command stream from the machine tool coordinate system to the workpiece coordinate system, and the tool axis vector and translational axis position coordinates are calculated. Parametric curve interpolation is used to perform trajectory smoothing fitting on the tool axis vector and translational axis position coordinates to construct a continuous tool center point motion trajectory and tool axis attitude change surface. According to the equal arc length discretization criterion, the continuous tool center point motion trajectory and attitude change surface are synchronously sampled. The tangential vector and normal vector at each sampling point are calculated, and the spatial position data and attitude data are time-aligned and encapsulated to generate a discrete tool position sequence.

3. The automatic optimization method for machining parameters of a five-axis CNC machine tool according to claim 1, characterized in that, The specific implementation process of using geometric Boolean operations to perform time-domain material removal on the discrete tool position sequence to generate time-varying geometric features containing the gradient of instantaneous contact area and material removal rate includes: Based on the temporal logic of the discrete tool position sequence, a sweep volume envelope of the tool along the motion trajectory is constructed. A solid construction geometry algorithm is used to perform Boolean difference operations between the sweep volume envelope and the current machining material, and the intermediate geometric state of the workpiece after material removal is updated in real time. The contact area boundary between the tool cutting edge and the intermediate geometric state of the workpiece at each discrete moment is extracted, and the instantaneous contact area between the tool and the workpiece is obtained by surface integration. The instantaneous material removal rate is calculated based on the volume increment of the removed material entity within adjacent time steps, and the sequence of the instantaneous material removal rate is subjected to temporal differentiation to extract the material removal rate gradient characterizing the abrupt change trend of the cutting force. The instantaneous contact area and the material removal rate gradient are associated and stored as time-varying geometric features.

4. The automatic optimization method for machining parameters of a five-axis CNC machine tool according to claim 1, characterized in that, The specific implementation process of constructing a CNC machine tool simulation model based on the material constitutive equation and frequency domain response function, and receiving the time-varying geometric features to calculate the regenerative chatter prediction sequence characterizing the dynamic stability of the machining process includes: The thermodynamic state of the cutting deformation zone is analyzed based on the physical properties of the processed material, generating a material constitutive equation related to the cutting speed and feed rate. A frequency domain response function containing time-varying delay terms is established by combining the instantaneous contact area in the time-varying geometric features. Based on the material constitutive equation and the frequency domain response function, a CNC machine tool simulation model describing the relative vibration between the tool and the processed material is constructed. The stability of the CNC machine tool simulation model is determined by a stability discrimination solution, generating a stable blade boundary characterizing the ultimate cutting depth. The time-varying geometric features are mapped to the space defined by the stable blade boundary, identifying frequency bands and locations exceeding the stability domain, and generating a regenerative chatter prediction sequence.

5. The automatic optimization method for machining parameters of a five-axis CNC machine tool according to claim 1, characterized in that, Based on the machine tool's current global stiffness and local flexibility data, the specific implementation process of using the regenerative chatter prediction sequence to dynamically constrain the feed rate within the control parameter space and delineate the real-time parameter feasible region that varies with the five-axis attitude includes: Based on the topology of the five-axis machine tool workspace, global stiffness data is invoked, and the instantaneous stiffness matrix under the current machining posture is calculated using a multi-dimensional spatial interpolation algorithm according to the translational axis position coordinates in the discrete tool position sequence. Local compliance data of the tool-workpiece contact interface is received and combined with the regenerative chatter prediction sequence to perform modal coupling correction on the instantaneous stiffness matrix. Within the control parameter space composed of spindle speed, feed rate, and acceleration, dynamic feed rate constraints to suppress regenerative chatter are generated. The continuous parameter set that satisfies the dynamic feed rate constraints is envelope-fitted to delineate the real-time parameter feasible region that changes with the five-axis posture.

6. The automatic optimization method for machining parameters of a five-axis CNC machine tool according to claim 1, characterized in that, Within the feasible region of the real-time parameters, the specific implementation process of using a multi-objective optimization algorithm to perform global iterative optimization of the set of machining parameter functions, including machining efficiency, surface contour error index, and machine tool energy efficiency ratio, includes: Machining efficiency is quantified as the volume of material removed per unit time. Surface contour error is defined as the weighted superposition of tool path approximation error and servo dynamic following error. Machine tool energy efficiency ratio is characterized as the energy conversion efficiency of effective cutting power and total drive input power. Within the feasible region of the real-time parameters, a set of machining parameter functions is generated using a random sampling strategy. Individuals in the set of machining parameter functions are hierarchically sorted according to the Pareto dominance relationship, and offspring populations are generated by selecting crossover and mutation operators. Whether the offspring population individuals cross the boundary of the feasible region of the real-time parameters is monitored in real time, and global iterative optimization is performed on the set of machining parameter functions.

7. The automatic optimization method for machining parameters of a five-axis CNC machine tool according to claim 1, characterized in that, The specific implementation process of automatically optimizing the machining process of a five-axis CNC machine tool by locking the machining parameter function set corresponding to the Pareto optimal solution set during the iterative optimization process includes: The Pareto front solution set output by the global iterative optimization is extracted. Combined with the current processing material's process preferences and production constraints, the utility of the non-dominated solutions in the Pareto front solution set is evaluated, and the processing parameter function set with the highest utility value is selected. The original CNC machining program file is read, and the original instruction stream is dynamically reconstructed and updated using the processing parameter function set to generate target CNC code containing optimized processing parameters. The target CNC code is converted into servo drive instructions by the machine tool control interpreter, controlling the servo motors of each axis to continue machining after adaptive parameter adjustments.

8. An automatic optimization system for machining parameters of a five-axis CNC machine tool, characterized in that, include: The parameter processing module performs discrete analysis on the machine tool parameters, including the tool axis vector and translation axis position coordinates, during the machining process to generate a discrete tool position sequence; it then uses geometric Boolean operations to perform time-domain material removal on the discrete tool position sequence to generate time-varying geometric features containing instantaneous contact area and material removal rate gradient. The machine tool control module constructs a CNC machine tool simulation model based on the material constitutive equation and frequency domain response function, receives the time-varying geometric features, and calculates a regenerative chatter prediction sequence characterizing the dynamic stability of the machining process. Based on the current global stiffness data and local flexibility data of the machine tool, it uses the regenerative chatter prediction sequence to perform dynamic feed speed constraints in the control parameter space and delineates the real-time parameter feasible region that varies with the five-axis attitude. The automatic optimization module uses a multi-objective optimization algorithm to perform global iterative optimization on the machining parameter function set, which includes machining efficiency, surface contour error index and machine tool energy efficiency ratio, within the feasible domain of the real-time parameters; it locks the machining parameter function set corresponding to the Pareto optimal solution set during the iterative optimization process, and automatically optimizes the machining process of the five-axis CNC machine tool.

Images