Honeycomb core six-axis ultrasonic cutting track digital control and interpolation optimization system
The six-axis ultrasonic cutting trajectory digital control and interpolation optimization system for honeycomb cores solves the problems of code expansion, speed instability, and insufficient dynamic parameter sensing in the cutting of complex free-form honeycomb cores, achieving efficient and stable cutting quality and equipment reliability.
Patent Information
- Application Number
- CN202610034098.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-01-12
- Publication Date
- 2026-04-10
AI Technical Summary
Existing technologies suffer from a dramatic increase in CNC code volume, unsmooth speed planning, and high real-time computational pressure in cutting complex free-form honeycomb cores. Furthermore, they lack the ability to perceive and adaptively compensate for dynamic parameters online, which affects processing quality and equipment lifespan.
A six-axis ultrasonic cutting trajectory digital control and interpolation optimization system with honeycomb core is adopted, which includes trajectory preprocessing, parametric spline interpolation, six-axis linkage inverse kinematics, dynamic feedforward compensation, ultrasonic energy adaptive control and multi-source information fusion module, forming a closed-loop system to realize trajectory planning, interpolation calculation, motion control and process adaptive optimization.
It significantly reduces the amount of CNC code, improves the surface finish of the cut surface and tool life, enhances the stability and consistency of machining quality, and reduces overall machining costs.
Smart Images

Figure CN121832449A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The application belongs to the field of numerical control machining and computer-aided manufacturing software, and particularly relates to a honeycomb core six-axis ultrasonic cutting trajectory digital control and interpolation optimization system. BACKGROUND
[0002] In the high-end manufacturing field of aerospace, rail transportation, etc., composite honeycomb cores are widely used in the manufacturing of lightweight structural components due to their excellent specific strength, specific stiffness and damping performance. The precise cutting of honeycomb cores is a key pre-process for component forming, and the machining quality directly affects the performance and assembly accuracy of the final product, thus putting forward very high requirements for the control accuracy and processing efficiency of the cutting trajectory.
[0003] The ultrasonic vibration cutting technology based on multi-axis numerical control system is the mainstream method to realize high-efficiency and precision machining of complex curved surface honeycomb cores. This technology integrates an ultrasonic vibration tool head with a multi-axis linkage numerical control platform, aiming to realize high-precision following of the preset three-dimensional cutting trajectory and high-quality removal of materials.
[0004] The existing technology usually adopts traditional numerical control interpolation algorithms, such as linear interpolation or circular arc interpolation, to generate multi-axis linkage instructions to drive the ultrasonic cutting tool. However, when facing the common cutting task of complex free curved surfaces in aerospace components, the existing method faces significant technical bottlenecks: in order to meet the high-precision contour requirements, the traditional interpolation algorithm needs to discretize the complex curve into a large number of small line segments, resulting in a dramatic increase in the amount of numerical control code, which puts a huge pressure on the real-time computing and data transmission capabilities of the control system; at the same time, the frequent speed planning between small line segments is difficult to smooth, which easily causes machine vibration, affecting the cutting surface quality and tool life. In addition, the existing system lacks the ability of online sensing and compensation of the dynamic parameters of the cutting process, and cannot adaptively adjust the interpolation parameters to cope with the machining disturbance caused by the local characteristics of the material. SUMMARY
[0005] The purpose of the present application is to provide a honeycomb core six-axis ultrasonic cutting trajectory digital control and interpolation optimization system to solve the technical contradictions in the prior art caused by the use of traditional interpolation algorithms to process complex free curved surface cutting trajectories, such as the dramatic increase in the amount of numerical control code, the non-smooth speed planning, the high real-time computing pressure of the system, and the lack of online sensing and adaptive compensation ability of the dynamic parameters of the machining process.
[0006] To achieve the above-mentioned purpose, the present application provides a honeycomb core six-axis ultrasonic cutting trajectory digital control and interpolation optimization system. The system includes a trajectory preprocessing module, a parameterized spline interpolation module, a six-axis linkage inverse solution module, a dynamic feedforward compensation module, an ultrasonic energy adaptive control module, and a multi-source information fusion and decision module. These modules work together to form a complete closed-loop system from trajectory planning, interpolation calculation, motion control to process adaptive optimization.
[0007] The trajectory preprocessing module is used to receive and process the input cellular core three-dimensional cutting trajectory data. The module first performs adaptive density sampling based on the rate of curvature change on the original trajectory point cloud, reduces the sampling density in the area where the curvature changes gently, and increases the sampling density in the area where the curvature changes sharply, thereby significantly reducing the number of discrete data points for subsequent processing while ensuring the accuracy of the profile. Further, the module has a trajectory smoothing filter based on the least square method built in, which is used to filter out high-frequency jitter introduced in the original data due to measurement noise or modeling errors, and generate a smooth initial trajectory point sequence.
[0008] The parameterized spline interpolation module is connected to the trajectory preprocessing module, and is used to reconstruct the preprocessed discrete trajectory point sequence into a continuous differentiable path. The module uses a 5th order non-uniform rational B-spline curve as the core interpolation tool. The specific implementation process is as follows: taking the preprocessed trajectory points as the control points, the corresponding node vectors are calculated by the cumulative chord length method; then, the control points are calculated based on the boundary condition constraints and the control points, and a global C2 continuous 5th order non-uniform rational B-spline curve is constructed. The curve can accurately fit various geometric shapes including straight lines, circular arcs and complex free curves. The interpolation module calculates the coordinates of the corresponding parameter points on the spline curve, the first derivative, i.e. the tangential velocity vector, and the second derivative, i.e. the curvature vector, according to the interpolation period and the feed speed set by the system, to provide continuous and high-order smooth reference position instructions for six-axis linkage.
[0009] The six-axis linkage inverse solution module is connected to the parameterized spline interpolation module, and is used to solve the actual position instructions of driving each motion axis of the six-axis numerical control platform from the tool center point space motion instructions described by the spline curve. The module has a machine tool kinematics model built in, which accurately describes the pose transformation relationship of the ultrasonic cutting tool head relative to the six-axis platform base. For the tool center point coordinates and attitude vector input for each interpolation period, the module calculates the target positions of the three linear axes and three rotary axes of the platform by solving the inverse kinematics. Further, the module integrates a singular pose avoidance algorithm based on the Jacobian matrix. When it is detected that the current or adjacent pose may cause the Jacobian matrix to tend to be ill-conditioned, the tool attitude planning is automatically fine-tuned to ensure the stability and continuity of the motion solution.
[0010] The dynamics feedforward compensation module is connected to the six-axis linkage inverse solution module, and is used to predict and compensate the tracking error of each motion axis caused by system inertia, friction and stiffness when following high-speed and high-acceleration instructions. The module first obtains the accurate dynamics model of each motion axis servo system through offline or online identification, including mass, damping, stiffness coefficient and Coulomb friction and viscous friction parameters. In each control cycle, the module receives the instruction position, speed and acceleration sequence of each axis from the six-axis linkage inverse solution module, and based on the identified dynamics model, forward calculates the theoretical driving torque or current required by each axis to accurately track the instruction. The calculated feedforward compensation is superimposed on the position loop or speed loop output of each axis servo driver, thereby significantly reducing the following error of each axis at high-speed contour motion, and improving the overall trajectory tracking accuracy.
[0011] The ultrasonic energy adaptive control module is independent of the trajectory interpolation and motion control loop, and is specially used to dynamically adjust the output energy of the ultrasonic vibration tool head according to the real-time cutting state. The input end of the module is connected to a multi-sensor array, which at least includes a three-dimensional force sensor for monitoring the contact force between the tool head and the honeycomb core, an amplitude sensor for monitoring the vibration amplitude of the tool head, and an infrared temperature sensor for monitoring the temperature of the tool head. The core of the module is a fuzzy inference-based energy regulator. The input variables of the energy regulator are the normalized real-time contact force deviation, amplitude deviation and temperature value; the output variable is the adjustment amount of the ultrasonic generator output power. The energy regulator internally presets a set of fuzzy rule library optimized through process experiments, for example, if the contact force is too large and the amplitude is too small, the output power is increased by a large amplitude. Through real-time defuzzification calculation, the module outputs accurate power adjustment instructions, ensuring that the cutting process maintains constant effective cutting energy in different material density areas or tool wear states, thereby ensuring consistent cutting surface quality and prolonging tool life.
[0012] The multi-source information fusion and decision module serves as the top-level coordination and optimization hub of the system, and performs bidirectional data interaction with all the above-mentioned modules. The module collects in real time the current curvature and feed speed information from the parameterized spline interpolation module, the actual position and command position error of each axis from the six-axis linkage inverse solution module, the load rate estimates of each axis from the dynamics feedforward compensation module, and the sensor readings and power adjustment state from the ultrasonic energy adaptive control module. The module has an online interpolation parameter optimizer based on the model predictive control principle. The optimizer is based on system state prediction in a future finite time domain, has the optimization objectives of minimizing the trajectory tracking error, load balancing of each axis, and cutting energy fluctuation, and has the decision variables of feed speed and interpolation period fine adjustment amount. A quadratic programming problem with constraints is solved in each optimization period. The optimal feed speed adjustment command obtained by the solution is fed back to the parameterized spline interpolation module to dynamically adjust the interpolation speed. Meanwhile, the generated system health status evaluation signal can be used to trigger early warning or maintenance prompts.
[0013] As an embodiment of the application, the parameterized spline interpolation module further integrates a forward-looking speed planning unit. The unit performs forward-looking speed planning on the entire or segmented spline curve based on the curvature information of the spline curve, the speed, acceleration, and jerk physical limits of each motion axis. The planning process follows an S-shaped acceleration / deceleration curve to ensure smooth speed change without impact. The unit can identify high-curvature corner regions in the curve and decelerate in advance, and then smoothly accelerate to the set feed speed after passing through the corner, thereby maximizing the average machining efficiency while ensuring contour accuracy.
[0014] As an embodiment of the application, the online identification function in the dynamics feedforward compensation module is implemented by the recursive least squares method. During system idling or specific calibration motion, the module injects a white noise test signal with controllable amplitude into each axis servo driver, and collects the actual position response of each axis. The input and output data stream collected is processed in real time by the recursive least squares method, and the parameter estimates of the dynamics model of each axis are constantly updated and corrected, so that the feedforward compensation can be adapted to slow changes in the mechanical properties of the system, such as wear and lubrication state changes.
[0015] As an embodiment of the application, the model predictive control optimizer in the multi-source information fusion and decision module has a simplified linear time-varying system as the system state prediction model, which includes the second-order dynamics characteristics of each axis and the discrete-time model of the interpolator. The weight coefficients of the optimization objectives can be preconfigured according to different types of honeycomb core materials and cutting quality grade requirements. The constraint conditions of the decision variables are directly derived from the performance specification sheets of the servo drivers of the numerical control platform.
[0016] As an embodiment of the present application, the whole system runs on a deterministic real-time operating system, and each functional module exists in the form of a periodic task with different priorities. The trajectory preprocessing module and the parametric spline interpolation module are high-priority tasks, ensuring the real-time performance of trajectory generation and interpolation calculation; the six-axis linkage inverse solution module and the dynamics feedforward compensation module are medium-priority tasks, ensuring the timeliness of motion instruction calculation and compensation; the ultrasonic energy adaptive control module and the multi-source information fusion and decision module are low-priority tasks, performing relatively slow process optimization and decision functions. Data exchange between modules is realized through shared memory and real-time message queues, ensuring the determinism and low latency of system response.
[0017] Compared with the prior art, the present application has the following advantages: 1. By introducing a parametric spline interpolation module, the present application uses a 5th-order non-uniform rational B-spline curve to globally and continuously fit complex cutting trajectories, fundamentally replacing the traditional method of discrete approximation by small line segments. This technical solution can describe complex curved surface trajectories using a small number of control points and a compact mathematical expression, greatly reducing the amount of generated NC code, and greatly reducing the storage and real-time calculation burden of the control system. At the same time, the inherent high-order continuity of the spline curve ensures smooth transitions in position, speed and even acceleration of the interpolation points, effectively avoiding machine vibration caused by frequent start-stop or sudden speed changes, thereby significantly improving the smoothness of the cutting surface and the service life of the tool.
[0018] 2. The present application constructs a closed-loop optimization system including dynamics feedforward compensation and multi-source information fusion and decision. The dynamics feedforward compensation module actively offsets tracking errors caused by inertia, friction and other factors of the motion axes based on an accurate system model, improving the contour accuracy at high speed and high precision. The multi-source information fusion and decision module coordinates interpolation parameters and system states online through a model predictive control algorithm, achieving dynamic optimization of the machining process. This system-level design makes the present application not only a trajectory following system, but also an intelligent machining system with online perception, modeling, decision-making and execution capabilities, capable of adaptively responding to material property changes, tool wear and other machining disturbances, ensuring the stability and consistency of the cutting quality.
[0019] 3, The application creatively incorporates ultrasonic energy control into the overall trajectory control closed loop, and realizes real-time accurate regulation of cutting energy through an ultrasonic energy adaptive control module. Based on multi-sensor information such as force, amplitude, and temperature, the module uses fuzzy reasoning intelligent decision output power to ensure constant and optimal cutting energy under different processing conditions. This technical solution solves the problem of fixed energy output in traditional ultrasonic cutting, which cannot adapt to changes in working conditions. Not only does it optimize cutting quality and reduce material damage, but it also effectively protects the ultrasonic tool head by preventing excess or insufficient energy, further extending the service life of key components and reducing overall processing costs. BRIEF DESCRIPTION OF DRAWINGS
[0020] Figure 1 is the overall technical scheme architecture schematic diagram of the application; Figure 2 is the core principle framework schematic diagram of the parameterized spline interpolation module in the application; Figure 3 is the logic flow framework diagram of the multi-source information fusion and decision module in the application; Figure 4 is the multi-level interaction relationship and data flow schematic diagram of the six-axis linkage inverse solution module and the dynamics feedforward compensation module in the application; Figure 5 is the core principle framework schematic diagram of the ultrasonic energy adaptive control module in the application. DETAILED DESCRIPTION
[0021] Embodiment 1: The honeycomb core six-axis ultrasonic cutting trajectory digital control and interpolation optimization system described in the application has the overall architecture as shown in the accompanying Figure 1 , which is composed of six functional units: trajectory preprocessing module, parameterized spline interpolation module, six-axis linkage inverse solution module, dynamics feedforward compensation module, ultrasonic energy adaptive control module, and multi-source information fusion and decision module. The modules interact with each other through a deterministic real-time communication mechanism to form a complete closed-loop control system from trajectory input, path reconstruction, motion calculation, dynamics compensation to process adaptive adjustment. Please refer to Figure 1 to Figure 5 , the entire system is deployed on an embedded computing platform with hard real-time characteristics and runs in a real-time operating system kernel environment that supports priority preemption and time-triggered scheduling, ensuring that critical control tasks are completed under strict timing constraints.
[0022] First, the trajectory preprocessing module, serving as the system's front-end data entry point, receives externally imported 3D clipped trajectory point cloud data for the honeycomb core. This point cloud data typically originates from discrete coordinate sequences exported by computer-aided design software, often exhibiting non-uniform sampling density and potentially containing high-frequency disturbances introduced by modeling errors or measurement noise. To address this issue, the trajectory preprocessing module first executes an adaptive resampling algorithm based on the rate of curvature change. This algorithm traverses the original trajectory point sequence, calculating the local curvature and its first-order difference (i.e., the rate of curvature change) for each point. When the rate of curvature change of a certain trajectory segment is lower than a preset threshold of 0.05mm... -1 ·mm -1 When the curvature change rate is higher than 0.2 mm, it is considered a flat region, and an interval skipping strategy is adopted, retaining one of every 3 to 5 original points; -1 ·mm -1 When a high-curvature corner or complex transition zone is identified, a local encryption mechanism is activated, inserting one or two new points between adjacent points. The insertion positions are obtained through cubic spline interpolation. This process compresses the original trajectory of tens of thousands of points to a scale of several thousand points while ensuring that the contour geometric features are not distorted. Subsequently, the module calls the built-in least-squares trajectory smoothing filter. This filter constrains the square integral of the second derivative of the trajectory with a quadratic regularization term, constructing the following optimization objective: minimizing the fitting error while suppressing trajectory jitter. The filter window length is set to 7 points, and the weight coefficient is experimentally calibrated to 0.8 to ensure effective filtering of high-frequency noise while preserving sharp features. The final output is a set of smooth, low-redundancy, and high geometrically faithful initial trajectory point sequences, which serve as the input reference for subsequent interpolation modules.
[0023] Please refer to the attached document. Figure 2 After receiving the discrete point sequence from the trajectory preprocessing module, the parameterized spline interpolation module initiates five non-uniform rational interpolation iterations. Spline curve reconstruction process. This process first uses the cumulative chord length method to calculate the parameter values corresponding to the model points. Specifically, let the initially set start endpoint parameters... =0, tail endpoint parameter =1, intermediate point parameter ,in The parameter value corresponding to the (i-1)th type value point. The total chord length of the entire trajectory, For the first Coordinates of the point of each type value For the first The coordinates of the point values are obtained. After obtaining the parameter sequence, the node vector is constructed. ,in These are B-spline values 0, 1, ... The node vector, , For the spline degree (here) =5), The number of type value points is reduced by 1, and the node vectors adopt a quasi-uniform distribution strategy, repeating at the beginning and end. +1 times to satisfy the endpoint interpolation condition. Then, based on the given boundary conditions (usually setting the first derivative at the first and last points as the zero vector, i.e., the natural boundary conditions), a system of linear equations is established to solve for the control point sequence. The system of equations is in the form of · = ,in For the reason spline basis functions in parameters The value formed by taking the position ( +1)×( +1) Coefficient matrix, This is a column vector of point values. Because it uses 5th-order splines and satisfies the C2 continuity requirement, the resulting curve has continuous position, velocity, and acceleration information globally. During system operation, the interpolation module operates at a fixed period... (Typical value is 1 millisecond) Trigger interpolation calculation. Within each cycle, the module calculates the interpolation based on the current accumulated arc length. and setting the feed rate The parameters are solved using the Newton-Raphson iterative method, so that the curve is in The arc length at that point is Then, the deBoor recursive algorithm is called to calculate the... The curve position corresponding to the value First derivative (i.e., tangential velocity vector) and second derivative (Used for curvature calculation). Curvature The result is calculated using the formula: ; This curvature value is transmitted in real time to the multi-source information fusion and decision-making module for velocity planning and load prediction. The interpolation module outputs... , , The complete spatial motion command for the tool center point is formed, including position and attitude information (the attitude is composed of the Frenet frame formed by the tangential, normal and subnormal directions), and sent to the six-axis linkage inverse kinematics module.
[0024] Combined with appendix Figure 4, the six-axis linkage inverse solution module receives the tool center point pose command from the parametric spline interpolation module, and performs inverse kinematics solution according to the pre-stored machine tool kinematics model. The machine tool is a typical 6-DOF serial structure, which includes 3 linear axes (X, Y, Z) and 3 rotary axes (A, B, C), and its forward kinematics is represented by a series of homogeneous transformation matrix multiplications. The inverse solution process adopts a hybrid strategy combining analytical method and numerical iteration: for the substructure with closed-form solution (such as the 3R mechanism orthogonal to the wrist), the rotary axis angles are directly solved using the analytical formula; for the linear-rotary combined part with coupling, the Newton-Raphson iteration method based on the Jacobian matrix is adopted. In each iteration, the end pose error is calculated under the current joint angle configuration , and the joint angle increment is updated through the pseudo-inverse Jacobian matrix . To avoid numerical instability caused by singular configurations, a singularity detection unit is built into the module. This unit calculates the condition number of the Jacobian matrix in real time , and determines that the system is approaching a singular state when . At this time, the module starts the pose fine-tuning mechanism: under the premise of keeping the tool center point position unchanged, a new pose that minimizes is searched within the small rotary freedom (usually ±2° range around the tangential axis) along the tool axis. This new pose replaces the original command input to the inverse solver, ensuring the existence and numerical stability of the solution. The final output is the target position command of the 6 motion axes , with units of millimeters and degrees, and an accuracy of 0.001 millimeters and 0.001 degrees.
[0025] The dynamics feedforward compensation module is followed by the six-axis linkage inverse solution module, and its core function is to generate feedforward compensation signals to offset the dynamic tracking errors of each axis servo system. This module first relies on the accurate dynamics model of each motion axis. For any linear axis i, its dynamics equation can be expressed as: ; where is the equivalent mass, is the damping coefficient, is the stiffness coefficient, is the Coulomb friction amplitude, is the viscous friction coefficient, is the driver output torque, is the displacement of the linear axis , is the first derivative of displacement, i.e. velocity; is the second derivative of displacement, i.e. acceleration, is the sign function. These parameters are obtained through offline calibration or online identification. During system initialization, the module performs an open-loop frequency sweep test, recording the response of each axis under different frequency sinusoidal excitation, and uses the system identification toolbox to fit the initial parameter set. During the machining process, the module enables recursive least squares to achieve online parameter update. Specifically, in each control cycle, a white noise disturbance signal with an amplitude of 5% of the rated current is injected into the servo driver, and the actual , velocity , acceleration and drive current (proportional to ) of the axis are synchronously collected. The dynamic equation is rewritten in the form of linear regression: , where the regression vector and the parameter vector . The recursive least squares method gives exponential decay weights to historical data with a forgetting factor =0.995, and updates the estimate in real time. After obtaining the latest model parameters, the module calculates the theoretical required drive torque according to the command trajectory of each axis and its first and second derivatives output by the six-axis linkage inverse solution module. After coordinate transformation and unit conversion, this feedforward torque is converted into a current command , which is superimposed on the output of the servo driver's speed loop. This compensation mechanism significantly reduces the tracking error of the position loop, especially at high speed corners or high acceleration sections, reducing the error.
[0026] The ultrasonic energy adaptive control module operates independently but is deeply coupled with the main motion control loop. Please refer to the attached Figure 5 , the module senses the cutting state in real time through a multi-sensor array. A three-dimensional force sensor is installed between the ultrasonic tool head and the main shaft connection flange, with a sampling frequency of 1 kHz, outputting X, Y, Z three-direction contact force . The amplitude sensor is a non-contact laser displacement meter, focused on the tool head end face, measuring its vibration peak-to-peak value , with a typical working range of 10 to 50 microns. The infrared temperature sensor is aimed at the tool head edge area, with a sampling frequency of 100 Hz, measuring a temperature range of 20 to 200 degrees Celsius. All raw signals are sent to the signal conditioning unit inside the module after anti-aliasing filtering and 16-bit analog-to-digital conversion. The unit first calculates the combined contact force , and compares it with the preset target force (typically 8 Newton), to obtain the force deviation . The amplitude deviation , where Set a reference amplitude for the process (e.g. 30 microns). The temperature value T is directly taken as the third input variable. The 3 input variables are normalized to the interval [-1, 1]: , is the normalized value of the force deviation, is the normalized value of the amplitude deviation, is the normalized value of the temperature, is the temperature. The maximum range of the amplitude = 5 N, the maximum range of the amplitude = 10 microns, the minimum value of the temperature = 20 °C, the maximum value of the temperature = 150 °C. The normalized variables are input to the fuzzy inference engine. The engine uses a Mamdani-type structure, the input membership functions are a combination of triangles and trapezoids, and the output membership function is a singleton. The fuzzy rule base contains 27 rules, covering all combinations of the five linguistic values of the input variables: negative large, negative small, zero, positive small, and positive large.
[0027] The defuzzification uses the center of gravity method, and the output is the power adjustment amount of the ultrasonic generator. The adjustment amount is updated at a frequency of 100 Hz, and controls the generator output through an analog voltage signal (0 to 10 volts). This closed-loop regulation ensures that the effective cutting energy remains constant when there are sudden changes in the density of the honeycomb core material (e.g. from 50 kg / m3 to 150 kg / m3) or slight wear of the tool, and the roughness Ra of the cut surface is stabilized within 12 microns.
[0028] The multi-source information fusion and decision module serves as the top-level intelligent hub of the system. Please refer to the attached Figure 3 , which has a logic flow divided into four stages: state perception, prediction modeling, optimization solving, and instruction issuance. In the state perception stage, the module polls each sub-module at a 10 millisecond cycle: it obtains the current curvature and feed speed from the parameterized spline interpolation module; it reads the command positions of each axis and the encoder feedback positions from the six-axis linkage inverse solution module, calculates the tracking error ; it extracts the load rates of each axis from the dynamics feedforward compensation module; it obtains , , and All data is timestamped and stored in a state buffer. During the predictive modeling phase, the module constructs a linear time-varying predictive model. This model discretizes the interpolator into a z-domain transfer function and simplifies each axis to a second-order inertial element. The overall system state vector x contains the position, velocity, and interpolation phase of each axis. The state prediction for the next N=20 steps (i.e., 200 milliseconds) is achieved through state-space equations. It is obtained recursively, where for The state vector at time t, for The state vector at time t, for The state transition matrix at time t, for The input matrix at time step, The decision variable is the feed rate adjustment. In the optimization and solution phase, the module establishes a quadratic programming problem: the objective function... ,in , , These are the weighting coefficients, corresponding to trajectory accuracy, load balancing, and energy fluctuation, respectively. for The total number of items. for Time of the first Load rate of each axis. for Real-time load rate tracking error. for The power adjustment of the ultrasonic generator at any given time. Constraints include: the first... The absolute value of the feed rate adjustment at any moment (Speed adjustment limit), the first Time of the first The absolute value of the acceleration along each axis (Acceleration limit), the first Time of the first The absolute value of the acceleration (jump) on each axis (Jerk speed limit). This QP problem was solved in 5 milliseconds by an interior-point solver. Optimal solution. The data is sent to the parametric spline interpolation module for dynamic correction of the feed rate for the next cycle. Simultaneously, if the load rate of any axis exceeds 90% for 1 second, or the temperature exceeds 180°C, the module generates a warning signal, prompting the operator to check the tool or cooling status via the human-machine interface.
[0029] As a system-level implementation detail, all the aforementioned modules are deployed as periodic real-time tasks. The trajectory preprocessing module runs only once before machining begins, typically taking less than 500 milliseconds. The parametric spline interpolation module and the six-axis linkage inverse kinematics module run at 1 kHz, assigned the highest CPU priority; the dynamic feedforward compensation module runs at the same frequency, with the next highest priority; the ultrasonic energy adaptive control module runs at 100 Hz, and the multi-source information fusion and decision-making module runs at 100 Hz, sharing a lower priority. Data exchange between modules is achieved through double-buffered shared memory, with locks applied to write operations and no locks applied to read operations to ensure data consistency. The communication latency has been measured to be less than 50 microseconds. Upon system startup, the machine tool kinematic parameters, initial values of the dynamic model, fuzzy rule base, and MPC weight configuration file are loaded first. The configuration file can be automatically switched according to the type of honeycomb core material (such as Nomex, aluminum honeycomb, and glass fiber). For example, when machining high-density aluminum honeycomb, the weight is increased by 30% to more strictly control energy fluctuations.
[0030] Furthermore, the look-ahead velocity planning unit integrated into the parametric spline interpolation module is activated immediately after the spline curve is generated. This unit traverses the entire curve, calculating the curvature at each point. And based on the speed limits of each axis Acceleration limit and accelerometer limit Calculate the maximum allowable feed rate at that point. . Constrained by the following formula: ,in For the first The equivalent curvature components corresponding to the axis are then used for S-shaped velocity planning. A forward scan determines the end point of the acceleration phase, and a reverse scan determines the start point of the deceleration phase, forming a smooth velocity groove in the corner region. The planning results are stored as a velocity-arc length lookup table, and the real-time feed rate is obtained during interpolation by looking up the table and using linear interpolation.
[0031] In summary, this embodiment constructs a high-precision, high-efficiency, and highly robust ultrasonic cutting control system for honeycomb cores through four core technologies: high-order spline interpolation, dynamic feedforward, ultrasonic energy adaptation, and multi-source fusion decision-making. This system not only solves the problems of code expansion and speed instability in traditional CNC methods when handling complex curved surfaces, but also actively suppresses disturbances in the processing through multi-level closed-loop feedback, ensuring consistent cutting quality and reliable equipment operation.
[0032] Example 2: Building upon Example 1, this example expands the optimization strategy of the multi-source information fusion and decision-making module for the scenario of regional cutting of large honeycomb core panels, introducing a partitioned collaborative scheduling mechanism. When the area covered by the cutting trajectory exceeds the effective travel of a single clamping (e.g., X-axis travel greater than 2000 mm), the system automatically divides the overall trajectory into several sub-regions, each defined by an independent local coordinate system. The multi-source information fusion and decision-making module adds a regional handover coordination function in this mode. At the boundary between adjacent sub-regions, the module initiates the handover protocol 500 mm in advance: First, the boundary segment trajectory (200 mm in length) is spline-fitted in both local coordinate systems to ensure seamless C2 continuity in the global coordinate system; second, 100 mm before the boundary, the phase and velocity of the interpolators on both sides are forcibly synchronized to eliminate instantaneous jumps introduced by coordinate system switching; finally, after the handover is completed, the actual tracking error of the boundary segment is statistically analyzed. If the root mean square error exceeds 5 micrometers, local trajectory replanning is triggered, and the position of the boundary control point is fine-tuned. This mechanism effectively solves the problem of misalignment of joints in the segmented processing of large workpieces.
[0033] Meanwhile, in this embodiment, the ultrasonic energy adaptive control module adds a material density online identification subunit. This subunit utilizes contact force. With feed rate steady-state relationship (in For material density, , (This is a process constant), collected during the linear uniform velocity segment. and The data is used to estimate the equivalent density of the current region in real time through least squares fitting. This density value is fed back to the fuzzy rule base selector, dynamically switching to a subset of rules that match that density range. For example, when When the concentration is <80 kg / m³, a low-density rule set should be used, as it is more sensitive to amplitude deviations; when... When the density is >120 kg / m³, a high-density rule set is activated to prioritize suppressing contact force overshoot. This adaptive mechanism enables the system to maintain stable cutting quality even when dealing with honeycomb core panels with density fluctuations of ±20% within the same batch.
[0034] Furthermore, the dynamic feedforward compensation module in this embodiment enhances the thermal deformation compensation function. During long-term continuous machining, the linear shaft guide rail undergoes thermal expansion due to frictional heat. The module constructs a temperature field model by using distributed temperature sensors (one per 500 mm) installed on each axis, combined with the material's thermal expansion coefficient. =12×10 -6 / ℃, calculate the thermal deformation at each location. . is the initial length of the corresponding position of the guide rail, is the temperature variation. The variation is added to the command position output by the six-axis linkage inverse solution module as an additional position offset. The compensation update frequency is 1 Hz, ensuring that the positioning error caused by thermal drift is controlled within 3 microns. The above improvements make the system particularly suitable for the high-precision integrated cutting needs of large honeycomb sandwich structure parts in the aerospace field.
[0035] Example 3: The honeycomb core six-axis ultrasonic cutting trajectory digital control and interpolation optimization system described in the present application, based on the infrastructure of example 1, adds a tool tip dynamic conversion unit and a tilt angle adaptive optimization unit to address small-angle cutting requirements and tool tip-workpiece contact adaptability issues. The control point is upgraded from the traditional tool center to the tool tip, forming a full-chain optimization system of "trajectory optimization-pose adaptation-energy matching". The entire system is still deployed on a hard real-time embedded computing platform and runs on a priority preemptive real-time operating system. The new units and the original modules work seamlessly together to ensure precision, efficiency, and surface quality improvement in small-angle cutting scenarios.
[0036] First, the trajectory preprocessing module adds an initial coordinate mapping function for tool tip points based on adaptive density sampling and smoothing filtering. After receiving the standard G-code tool path (including tool position point coordinates , tool orientation angle, etc.) based on the tool center, the module first performs adaptive resampling based on the curvature rate (1-2 new points are inserted in gentle areas and 1 new point is retained in severe areas), and then filters out high-frequency noise through a 7-point window, weight coefficient 0.8 least squares smoothing filter. Then, the tool tip coordinate conversion algorithm is called, and the tool tip length (preset 5mm) and the tool direction vector , are divided into , , axis vectors, and the initial tool tip coordinate is calculated by the formula to generate a smoothing trajectory point sequence centered on the tool tip.
[0037] The parametric spline interpolation module integrates a 3-5-3 polynomial interpolation trajectory smoothing unit based on 5th-order non-uniform rational B-spline curve reconstruction. The module uses the preprocessed tool tip points as the control point, calculates the parameter value by the cumulative chord length method, constructs the node vector, and solves the control point sequence to generate a globally C2 continuous curve. For small-angle cutting sections (forward tilt angle ≤15° or side tilt angle (≤10°), enable 3-5-3 polynomial interpolation: divide the path into three segments to ensure the continuity of displacement, velocity, and acceleration at the interpolation point, minimize acceleration jumps, and avoid machine tool vibration. The interpolation module outputs the tool edge coordinates and attitude information at a 1-millisecond cycle, which is synchronously transmitted to the deflection angle optimization unit and the six-axis linkage inverse kinematics module.
[0038] The newly added adaptive tilt angle optimization unit interacts bidirectionally with each module, dynamically adjusting the tool posture based on the surface curvature. Sampling points are extracted along the tool path (0.1mm interval in high curvature regions, 0.5mm interval in low curvature regions), and the tilt angle (forward tilt angle) is established using a radial basis function neural network (RBFN). yaw angle The mapping relationship between cutting angle and cutting quality is established. The optimal deviation angle is solved using the gradient ascent method and the tool attitude matrix. Adjust direction vector Update the blade position to ensure optimal contact between the blade and the honeycomb core surface.
[0039] The six-axis linkage inverse kinematics module adapts to the optimized pose commands for the cutting edge point and deflection angle, and solves the inverse kinematics using a hybrid strategy of "analytical method + numerical iteration". The singular configuration avoidance unit calculates the Jacobian matrix condition number in real time. hour, It is a Jacobian matrix The condition number is used to finely adjust the deflection angle while keeping the blade position unchanged, ensuring stable motion calculation, and finally outputting the target position commands for 6 motion axes (accuracy 0.001mm / 0.001°).
[0040] The dynamic feedforward compensation module achieves offline calibration and online identification (recursive least squares method, forgetting factor). =0.995) Update the dynamic model to adapt to the inertial characteristics of the blade point motion. Receive the instruction sequence from the inverse kinematics module, calculate the theoretical driving torque and convert it into current commands, which are then superimposed on the servo driver to enhance friction compensation on the rotating axis and reduce tracking errors during small-angle cutting.
[0041] The ultrasonic energy adaptive control module now includes an angle information input to optimize energy adjustment strategies. It collects triaxial contact force and tool head vibration peak-to-peak values at 1kHz and blade temperature at 100Hz. These data, combined with the normalized angle data, are then input into the fuzzy inference engine. An angle association rule has been added to the fuzzy rule base. After defuzzification, a power adjustment command is output at 100Hz to ensure energy matching with blade contact during small-angle cutting, maintaining a surface roughness Ra≤12μm.
[0042] The multi-source information fusion and decision module adds a reasonable evaluation of the deflection angle and monitoring of the contact stability of the cutting edge, and the optimization objective function is incorporated into the deflection angle weight coefficient (0.3 for small angle cutting). The data of each module is polled at a 10 millisecond cycle, and the optimal feed speed and deflection angle fine adjustment command is output by solving the quadratic programming problem through the interior point method; if the edge integrity is less than or equal to 90% or the layering degree is greater than or equal to 8%, a warning signal is generated.
[0043] At the system deployment level, the newly added deflection angle adaptive optimization unit runs at a frequency of 1 kHz (medium priority), and the 3-5-3 polynomial interpolation unit is triggered synchronously with the interpolation period. The modules interact through double-buffered shared memory, and the communication delay is less than or equal to 50 microseconds. The system supports presetting the deflection angle range according to the type of cellular core material (low-density material ≤20°, ≤15°; high-density material ≤30°, ≤20°), which adapts to the small angle cutting needs of different working conditions.
[0044] This embodiment upgrades the "dynamic conversion of cutting edge points + adaptive optimization of deflection angle + 3-5-3 polynomial interpolation" technology, realizes precise small angle cutting from 0° to 30°, and solves the technical pain points of traditional shaft control that cannot adapt to small angle cutting and poor surface quality.
Claims
1. A digital control and interpolation optimization system for six-axis ultrasonic cutting trajectory of honeycomb core, characterized in that, include: The trajectory preprocessing module is used to receive and process the input 3D cutting trajectory data of the honeycomb core; The parameterized spline interpolation module is connected to the trajectory preprocessing module and is used to reconstruct the preprocessed discrete trajectory point sequence into a continuous differentiable path. The six-axis linkage inverse kinematics module is connected to the parametric spline interpolation module and is used to convert the spatial motion command of the tool center point into the actual position command of each motion axis of the six-axis CNC platform. The dynamic feedforward compensation module is connected to the six-axis linkage inverse kinematics module to predict and compensate for the tracking error caused by system inertia, friction and stiffness when each motion axis follows high-speed and high-acceleration commands. An ultrasonic energy adaptive control module is used to dynamically adjust the output energy of the ultrasonic vibration tool head according to the real-time cutting status. The multi-source information fusion and decision-making module, as the top-level coordination and optimization hub of the system, interacts bidirectionally with all the above modules. This module collects in real time the current curvature and feed rate information from the parametric spline interpolation module, the actual position and command position error of each axis from the six-axis linkage inverse solution module, the estimated load rate of each axis from the dynamic feedforward compensation module, and the sensor readings and power adjustment status from the ultrasonic energy adaptive control module.
2. The digital control and interpolation optimization system for six-axis ultrasonic cutting trajectory of honeycomb core according to claim 1, characterized in that, The trajectory preprocessing module first performs adaptive density sampling on the original trajectory point cloud based on the rate of curvature change. The sampling density is reduced in regions with gentle curvature changes and increased in regions with drastic curvature changes, in order to reduce the number of discrete data points in subsequent processing while ensuring contour accuracy. The module has a built-in trajectory smoothing filter based on the least squares method to filter out high-frequency jitter in the original data and generate a smooth initial trajectory point sequence.
3. The digital control and interpolation optimization system for six-axis ultrasonic cutting trajectory of honeycomb core according to claim 2, characterized in that, The parameterized spline interpolation module uses a fifth-order non-uniform rational B-spline curve as the core interpolation tool. Its specific implementation process is as follows: using the pre-processed trajectory points as shape points, the corresponding node vectors are calculated by the cumulative chord length method. Then, based on the boundary condition constraints and the shape points, the control points are back-calculated to construct a globally C2 continuous fifth-order non-uniform rational B-spline curve. The interpolation module calculates the coordinates, first derivative (tangential velocity vector), and second derivative (curvature vector) of the corresponding parameter points on the spline curve in real time according to the interpolation period and feed rate set by the system.
4. The digital control and interpolation optimization system for six-axis ultrasonic cutting trajectory of honeycomb core according to claim 3, characterized in that, The six-axis linkage inverse kinematics module has a built-in machine tool kinematics model. For the tool center point coordinates and attitude vector input in each interpolation cycle, the module calculates the target positions of the platform's three linear axes and three rotary axes in real time by solving the inverse kinematics. The module integrates a singular configuration avoidance algorithm based on the Jacobian matrix. When it detects that the current or adjacent pose may cause the Jacobian matrix to become ill-conditioned, it automatically fine-tunes the tool attitude planning.
5. The digital control and interpolation optimization system for six-axis ultrasonic cutting trajectory of honeycomb core according to claim 4, characterized in that, The dynamic feedforward compensation module obtains the accurate dynamic model of each motion axis servo system through offline or online identification. In each control cycle, the module receives the command position, velocity and acceleration sequence of each axis from the six-axis linkage inverse kinematics module, and calculates the theoretical driving torque or current required for each axis to accurately track the command based on the identified dynamic model. The calculated feedforward compensation amount is superimposed on the position loop or velocity loop output of each axis servo driver.
6. The digital control and interpolation optimization system for six-axis ultrasonic cutting trajectory of honeycomb core according to claim 5, characterized in that, The input of the ultrasonic energy adaptive control module is connected to a multi-sensor array, which includes at least a three-dimensional force sensor for monitoring the contact force between the tool head and the honeycomb core, an amplitude sensor for monitoring the vibration amplitude of the tool head, and an infrared temperature sensor for monitoring the temperature of the tool head. The core of the module is an energy regulator based on fuzzy inference. The input variables of the energy regulator are the normalized real-time contact force deviation, amplitude deviation, and temperature value, and the output variable is the adjustment amount of the ultrasonic generator output power. The energy regulator has a pre-set set of fuzzy rule libraries optimized by process experiments.
7. The digital control and interpolation optimization system for six-axis ultrasonic cutting trajectory of honeycomb core according to claim 6, characterized in that, The multi-source information fusion and decision-making module incorporates an online interpolation parameter optimizer based on the model predictive control principle. This optimizer is based on the prediction of the system state in the future finite time domain, with the optimization objectives of minimizing trajectory tracking error, load balance of each axis, and cutting energy fluctuation. It uses feed rate and interpolation cycle fine-tuning as decision variables, and solves a constrained quadratic programming problem in each optimization cycle. The optimal feed rate adjustment command obtained from the solution is fed back to the parameterized spline interpolation module to dynamically adjust its interpolation rate.
8. The digital control and interpolation optimization system for six-axis ultrasonic cutting trajectory of honeycomb core according to claim 7, characterized in that, The parametric spline interpolation module also integrates a look-ahead velocity planning unit. Based on the curvature information of the spline curve, the velocity, acceleration, and physical limits of each motion axis, this unit performs look-ahead velocity planning for the entire or segmented spline curve. The planning process follows an S-shaped acceleration and deceleration curve. This unit can identify high curvature corner regions in the curve and decelerate in advance, then smoothly accelerate to the set feed rate after passing the corner.
9. The digital control and interpolation optimization system for six-axis ultrasonic cutting trajectory of honeycomb core according to claim 8, characterized in that, The online identification function in the dynamic feedforward compensation module is implemented by recursive least squares method. During system no-load or specific calibration motion, the module injects white noise test signals with controllable amplitude into each axis servo drive, and at the same time collects the actual position response of each axis. The recursive least squares method is used to process the collected input and output data streams in real time, and continuously update and correct the parameter estimates of each axis dynamic model.
10. The digital control and interpolation optimization system for six-axis ultrasonic cutting trajectory of honeycomb core according to claim 9, characterized in that, The model predictive control optimizer in the multi-source information fusion and decision module has a simplified linear time-varying system state prediction model that includes the second-order dynamic characteristics of each axis and the discrete-time model of the interpolator. The weight coefficients of the optimization objective can be preset according to different honeycomb core material types and cutting quality level requirements.
Citation Information
Cited By
A control system of an aviation composite ultrasonic seven-axis machining center
CN122143153A
