Intermittent cutting adaptive control system for preventing chipping of superhard blade

Abstract

The present application belongs to the technical field of numerical control machining, and discloses a discontinuous cutting self-adaptive control system for inhibiting the chipping of a superhard blade; comprising: a signal acquisition module, which is used for acquiring and processing original high-frequency vibration signals in the current discontinuous cutting process of the superhard blade to obtain cutting vibration time series data flow; a phase space reconstruction module, which is used for reconstructing the cutting vibration time series data flow based on an optimal delay time and an optimal embedding dimension to determine a high-dimensional phase space trajectory matrix; a chaotic feature analysis module, which is used for iteratively tracking phase points in the high-dimensional phase space trajectory matrix and determining a variable impedance control parameter when a chaos divergence precursor is identified; and a servo execution module, which is used for responding to the variable impedance control parameter instruction, adjusting the compensation torque of the feed shaft in real time, and completing the switching of the working mode of the superhard blade through the compensation torque, so as to improve the inhibition of chipping.

Description

Technical Field

[0001] This invention relates to the field of CNC machining technology, and more specifically, to an adaptive control system for intermittent cutting that suppresses chipping of superhard cutting inserts. Background Technology

[0002] With the increasing demand for processing high-hardness and difficult-to-machine materials in the high-end equipment manufacturing field, superhard cutting tools represented by PCBN and PCD have been widely used in various key industries due to their advantages of high hardness and high wear resistance. Interrupted cutting, as a common typical working condition in parts processing, is prone to triggering the most frequent chipping failure mode of superhard cutting tools due to its high frequency of mechanical impact, which seriously restricts processing efficiency and surface quality.

[0003] Existing technologies typically employ threshold triggering mechanisms to suppress chipping of superhard inserts. This involves using force sensors mounted on the machine tool or tool holder to collect cutting signals in real time, setting a fixed upper limit for the amplitude as an alarm threshold. When the detected signal amplitude exceeds this threshold, protective actions such as deceleration feed and rapid tool retraction are triggered. However, this passive linear threshold mechanism suffers from significant hysteresis. For superhard inserts, the extremely high brittleness of superhard materials results in a lack of a plastic buffer stage. The time from microcrack initiation to macroscopic chipping failure is extremely short, typically on the order of milliseconds. By the time the vibration amplitude detected by the sensor exceeds the linear threshold, the cutting edge of the superhard insert has often already undergone irreversible chipping. This prevents the control system from detecting and intervening in the early stages of chipping, making it difficult to meet the stringent requirements for timely early warning in high-speed intermittent cutting of superhard inserts.

[0004] In view of this, the present invention proposes an adaptive control system for intermittent cutting to suppress chipping of superhard cutting tools in order to solve the above problems. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art and to achieve the above objectives, the present invention provides the following technical solution: an adaptive control system for intermittent cutting to suppress chipping of superhard cutting inserts, comprising: The signal acquisition module is used to acquire and process the original high-frequency vibration signal during the current intermittent cutting process of the superhard cutting tool to obtain the cutting vibration time sequence data stream. The phase space reconstruction module is used to perform autocorrelation analysis on the cutting vibration time series data stream to determine the optimal delay time, and to determine the optimal embedding dimension through false nearest neighbor detection. Based on the optimal delay time and the optimal embedding dimension, the phase space of the cutting vibration time series data stream is reconstructed to determine the high-dimensional phase space trajectory matrix. The chaotic feature analysis module is used to iteratively track the phase points in the high-dimensional phase space trajectory matrix to obtain the maximum Lyapunov exponent. Based on the maximum Lyapunov exponent and its changing trend, it identifies whether the current intermittent cutting process is in the pre-chaotic divergence stage and determines the variable impedance control parameters when the pre-chaotic divergence stage is identified. The servo actuator module is used to respond to variable impedance control parameter commands, adjust the compensation torque of the feed axis in real time, and switch the working mode of the superhard cutting tool through the compensation torque.

[0006] Furthermore, the cutting vibration timing data stream is obtained, including: The cutting vibration of the superhard cutting tool during the current intermittent cutting process is acquired in real time, and the cutting vibration is converted into a raw high-frequency vibration signal that is proportional to the vibration acceleration based on the piezoelectric effect. The original high-frequency vibration signal is preprocessed to form a discrete digital signal sequence; the discrete digital signal sequence is arranged in time order to form a cutting vibration time-series data stream.

[0007] Further, autocorrelation analysis is performed to determine the optimal delay time, including: Set several candidate delay times, shift the cutting vibration time series data stream backward according to the step size of each candidate delay time and truncate it by the same length to obtain the delayed truncation sequence and the original truncation sequence under the corresponding candidate delay time; The interdependence between the original truncation sequence and the delayed truncation sequence under each candidate delay time is analyzed to determine the mutual information overlap under different candidate delay times. The mutual information overlap is then connected sequentially in ascending order of candidate delay time to form a mutual information overlap curve. The candidate delay time corresponding to the first time when the mutual information overlap reaches the valley point is selected from the mutual information overlap curve and determined as the optimal delay time.

[0008] Further, to determine the optimal embedding dimension, including: Set an increasing sequence of embedding dimensions, select one embedding dimension from the increasing sequence of embedding dimensions in turn, and reconstruct the cutting vibration time series data stream in combination with the optimal delay time to obtain the phase space state point set under each embedding dimension; False neighboring points are identified based on the rate of change of distance between each state point and its neighboring points in the state point set of each phase space as the dimension increases, and the proportion of false neighboring points to the number of neighboring points under each embedding dimension is counted. The embedding dimension corresponding to the first drop of the proportion of false neighboring points to a preset approach threshold is selected as the optimal embedding dimension.

[0009] Furthermore, the high-dimensional phase space trajectory matrix is ​​determined, including: Using the optimal delay time as the interval step, a number of state point row vectors with the optimal embedding dimension are sequentially extracted from the cutting vibration time series data stream as phase points; all phase points are arranged in time order to obtain a high-dimensional phase space trajectory matrix.

[0010] Furthermore, the phase points in the high-dimensional phase space trajectory matrix are traced, including: The first phase point in the high-dimensional phase space trajectory matrix is ​​selected as the initial reference phase point, and the phase point with the smallest Euclidean distance to the initial reference phase point is searched in the high-dimensional phase space trajectory matrix as the initial neighboring phase point; the initial neighboring distance between the initial reference phase point and the initial neighboring phase point is recorded. The initial reference phase point and the initial neighboring phase point are simultaneously shifted backward along the time axis by one evolution step to obtain the evolved reference phase point and the evolved neighboring phase point; Obtain the evolved neighbor distance, normalize and logarithmically transform the initial neighbor distance and the evolved neighbor distance to obtain the local divergence rate at the current evolution step size; The evolved reference phase point is used as the new reference phase point to iteratively track the trajectory matrix in the high-dimensional phase space until the reference phase point reaches the end of the trajectory matrix in the high-dimensional phase space, and the local divergence rate corresponding to each evolution step is determined.

[0011] Furthermore, the maximum Lyapunov index is obtained, including: An evolutionary relationship curve is constructed with the number of evolutionary steps as the x-axis and the local divergence rate as the y-axis; Inflection point detection is performed on the evolutionary relationship curve, and the continuous interval between the first inflection point and the second inflection point in the evolutionary relationship curve is determined as the linear growth region; The number of evolutionary steps and the corresponding local divergence rate within the linear growth region are fitted and analyzed. The slope of the straight line obtained from the fitting analysis is determined as the maximum Lyapunov exponent.

[0012] Furthermore, whether the current intermittent cutting process is in the early stages of chaotic divergence includes: The maximum Lyapunov exponents within the sliding analysis window are arranged in chronological order to form a historical evolution sequence; the linear fitting slope in the historical evolution sequence is extracted as the dynamic evolution trend of the discontinuous cutting process. The preset stability critical threshold and deterioration rate threshold are compared with the maximum Lyapunov exponent and dynamic evolution trend at the current moment, respectively. If the maximum Lyapunov exponent at the current moment is less than or equal to the stability critical threshold and the dynamic evolution trend is greater than the deterioration rate threshold, then the current discontinuous cutting process is determined to be in the precursor of chaotic divergence. In response to the identified precursors of chaotic divergence, an anti-impact parameter set matching the current cutting condition is extracted from a preset control parameter database; the anti-impact parameter set includes a high damping coefficient, a low stiffness coefficient, and a matching inertia coefficient.

[0013] Furthermore, the compensation torque of the feed axis is adjusted in real time, including: The theoretical cutting trajectory position of the superhard insert and the current actual position of the feed axis are obtained in the current interrupted cutting process, and the position tracking deviation is determined by differential comparison; based on the change of position tracking deviation over time, the current speed tracking error and acceleration tracking error of the feed axis are determined. By adjusting and compensating for position tracking deviation, velocity tracking error and acceleration tracking error using an anti-impact parameter set, a compensation torque containing inertia compensation component, damping compensation component and stiffness compensation component is obtained.

[0014] Furthermore, the switching of the superhard cutting tool's operating mode is accomplished through compensation torque, including: The compensation torque is converted into a current command and injected into the current control loop of the servo driver; the servo driver controls the feed axis to generate an electromagnetic torque corresponding to the compensation torque based on the current command. The electromagnetic torque drives the feed axis to generate a small displacement during the intermittent cutting impact of the superhard insert. Through this small displacement, the contact stiffness between the superhard insert and the workpiece is switched from the high-rigidity position holding mode under normal feed axis operation to the high-damping flexible buffer mode.

[0015] The technical effects and advantages of the intermittent cutting adaptive control system for suppressing chipping of superhard cutting inserts in this invention are as follows: 1. This invention achieves high-dimensional phase space mapping of cutting vibration data by relying on a phase space reconstruction module; by using the optimal delay time, it accurately captures the information decorrelation scale of discontinuous cutting in the nonlinear evolution process, thereby avoiding phase space trajectory distortion and information redundancy; by determining the optimal embedding dimension, it completely restores the cutting dynamic trajectory with the minimum dimension, so that the high-dimensional phase space trajectory matrix can truly reflect the nonlinear characteristics of discontinuous cutting, thus providing an accurate data base for chaotic feature analysis; 2. This invention uses a chaotic feature analysis module to iteratively track phase points in a high-dimensional phase space trajectory matrix. By constructing an evolutionary relationship curve and extracting the slope of the linear growth region as the maximum Lyapunov exponent, it quantifies the sensitivity of the intermittent cutting process to initial conditions. By fitting the dynamic evolution trend and combining the stability critical threshold and the deterioration rate threshold for dual-condition judgment, it can identify critical precursors that are accelerating towards a chaotic state in advance, even before the vibration amplitude shows obvious abnormalities and the chipping has actually occurred. 3. This invention generates a compensating torque in response to the detected chaotic divergence precursors via a servo execution module. This torque drives the feed axis to produce a minute displacement at the moment of intermittent cutting impact. At the moment of impact, vibration energy is dissipated by increasing damping and reducing stiffness, thus avoiding stress concentration caused by hard collision between the superhard insert and the workpiece, and effectively suppressing chipping. At the same time, by using only a minute displacement without tool retraction, the problem of low machining efficiency caused by frequent deceleration or tool retraction is avoided. This invention achieves active absorption of intermittent cutting impact energy without interrupting the cutting process, ultimately achieving a dual improvement in machining efficiency and tool life. Attached Figure Description

[0016] Figure 1 This is a system schematic diagram of the intermittent cutting adaptive control system for suppressing chipping of superhard cutting inserts according to the present invention. Figure 2 A flowchart for determining the optimal embedding dimension for this invention; Figure 3 This is a flowchart for determining whether a state is in the early stages of chaotic divergence, as described in this invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Example 1, please refer to Figure 1 , Figure 2 and Figure 3 As shown, the intermittent cutting adaptive control system for suppressing chipping of superhard cutting tools described in this embodiment includes: The signal acquisition module is used to acquire and process the original high-frequency vibration signal during the current intermittent cutting process of the superhard cutting tool to obtain the cutting vibration time sequence data stream.

[0019] The cutting vibration timing data stream is obtained, including: The cutting vibration of the superhard cutting tool during the current intermittent cutting process is acquired in real time, and the cutting vibration is converted into a raw high-frequency vibration signal that is proportional to the vibration acceleration based on the piezoelectric effect.

[0020] A triaxial piezoelectric accelerometer is rigidly attached to the tool holder surface adjacent to the superhard cutting insert. Utilizing the direct piezoelectric effect generated by the triaxial piezoelectric accelerometer under the impact pressure of the current intermittent cutting process (such as milling entry / exit, turning of workpieces with holes), the mechanical force generated by cutting vibration is converted into polarized charges based on the direct piezoelectric effect. These polarized charges are then fed into a charge amplifier with normalization function for impedance transformation and amplification, outputting a continuous analog voltage signal proportional to the vibration acceleration, which serves as the original high-frequency vibration signal.

[0021] The original high-frequency vibration signal is preprocessed to form a discrete digital signal sequence.

[0022] The original high-frequency vibration signal is transmitted to an analog-to-digital converter (ADC) via a shielded cable. An anti-aliasing low-pass filter at the front end of the ADC filters out high-frequency noise components exceeding the Nyquist frequency limit (e.g., setting the filter cutoff frequency to 0.4 times the system sampling frequency). A sample-and-hold circuit is triggered at a fixed high-frequency sampling frequency (e.g., 50kHz) to discretize the filtered original high-frequency vibration signal, quantizing the analog amplitude of each sampling point into a binary code, and outputting a set of voltage amplitude values ​​to form a discrete digital signal sequence.

[0023] Discrete digital signal sequences are arranged in chronological order to form a cutting vibration time-series data stream. Based on the sensitivity coefficient of the triaxial piezoelectric accelerometer (100mV / g), the voltage amplitudes in the discrete digital signal sequences are mapped point-to-point to vibration amplitudes characterizing the cutting impact intensity. The mapped vibration amplitudes are linearly sorted according to the timestamp index of the sampling time to form a one-dimensional vector containing time-domain information, thus forming the cutting vibration time-series data stream.

[0024] The phase space reconstruction module is used to perform autocorrelation analysis on the cutting vibration time series data stream to determine the optimal delay time, and to determine the optimal embedding dimension through false neighbor detection. Based on the optimal delay time and the optimal embedding dimension, the phase space of the cutting vibration time series data stream is reconstructed to determine the high-dimensional phase space trajectory matrix.

[0025] Perform autocorrelation analysis to determine the optimal delay time, including: Several candidate delay times are set. The cutting vibration time-series data stream is shifted backward according to the step size of each candidate delay time and truncated by the same length to obtain the delayed truncated sequence and the original truncated sequence for the corresponding candidate delay time. For each specific candidate delay time, the time axis of the cutting vibration time-series data stream is kept stationary, and a copy of the cutting vibration time-series data stream is made and shifted backward along the time axis by the candidate delay time sampling points. The original truncated sequence is obtained by truncating from the zeroth sampling point to the total length minus the delay time from the beginning of the original cutting vibration time-series data stream. The delayed truncated sequence is obtained by truncating from the delay time to the end of the total length from the end of the shifted cutting vibration time-series data stream. The delayed truncated sequence and the original truncated sequence correspond completely and are of the same length on the time axis.

[0026] It should be explained that the candidate delay time is set based on the physical impact cycle of the current intermittent cutting. Specifically, the actual spindle speed in the entire cutting system is read, and the time interval between two adjacent cutting impacts is calculated in combination with the number of teeth on the superhard insert, further obtaining the duration of a single impact cycle. An empirical rule is used to set the upper limit of the delay time search, which generally should not exceed one-third to one-half of the impact cycle. The impact cycle duration is multiplied by the sampling frequency to obtain the number of sampling points corresponding to the impact cycle. Then, this number of sampling points is divided by 3, and the resulting value is rounded down as the upper limit of the candidate delay time search (dividing by 3 ensures that the total embedding window length will not exceed one complete impact cycle when the minimum allowed embedding dimension of 2 is used for subsequent reconstruction).

[0027] The interdependence between the original and delayed truncation sequences at each candidate delay time is analyzed to determine the mutual information overlap at different candidate delay times. The vibration amplitudes of the original and delayed truncation sequences are discretized using an equidistant histogram method, forming a fixed number of equally spaced intervals (e.g., 60). The frequency of vibration amplitudes independently falling within each equally spaced interval is counted to calculate the marginal probability distribution. The frequency of vibration amplitudes of the original and delayed truncation sequences simultaneously falling within a specific interval combination (a two-dimensional grid cell formed by pairwise pairing of equally spaced intervals of the original and delayed truncation sequences) at the same time is counted to calculate the joint probability distribution. The total information between the original and delayed truncation sequences is calculated using the joint probability distribution and the marginal probability distribution (the total information is used to quantify the nonlinear correlation between the original and delayed sequences), and the calculated total information is determined as the mutual information overlap at that candidate delay time.

[0028] The mutual information overlap values ​​are sequentially connected in ascending order of candidate delay times to form a mutual information overlap curve. The candidate delay time corresponding to the first valley point of the mutual information overlap value is selected from the mutual information overlap curve and determined as the optimal delay time. The optimal delay time is used to ensure that the reconstructed coordinate axes maintain minimal information redundancy during the phase space reconstruction process. This allows the subsequently constructed high-dimensional phase space trajectory matrix to maximally represent the true structure of the cutting system's dynamic behavior, avoiding the obscuring of the system's nonlinear evolution characteristics due to coordinate axis information overlap.

[0029] All mutual information overlaps are arranged in ascending order of candidate delay times. Using candidate delay times as the x-axis and mutual information overlap as the y-axis, a spline interpolation algorithm is used to smoothly connect the mutual information overlaps, forming a mutual information overlap curve that reflects the decay of correlation over time. The mutual information overlap curve is then monotonically scanned to identify the first inflection point where the curve slope changes from negative to positive. This point is determined as the first local minimum of mutual information overlap, i.e., the valley point. The candidate delay time value pointed to by the valley point on the x-axis is extracted and determined as the optimal delay time.

[0030] To determine the optimal embedding dimension, including: Set an increasing embedding dimension sequence, and sequentially select an embedding dimension from the sequence. Combine this with the optimal delay time to reconstruct the cutting vibration time series data stream, obtaining the phase space state point set under each embedding dimension. The increasing embedding dimension sequence is a sequence with an initial value of 2 and an increment of 1 (e.g., setting the increasing embedding dimension sequence to a set of integers from 1 to 20).

[0031] For each current embedding dimension, with the optimal delay time as the fixed interval step, starting from the first vibration amplitude of the cutting vibration time-series data stream, a vector of length equal to the current embedding dimension is continuously truncated until no vector of complete length can be truncated. All truncated vectors are arranged in chronological order to form the phase space state point set under the current embedding dimension.

[0032] False neighbors are identified based on the rate of change of distance between each state point and its neighbors in each phase space state point set as the dimension increases. Each vector in the phase space state point set is considered a target state phase point. The vector closest to the target state phase point is searched within the phase space state point set and identified as the nearest neighbor phase point. The spatial distance between the target state phase point and the nearest neighbor phase point is calculated at the current embedding dimension, and the new spatial distance between them is also obtained when the dimension is increased to the next order of embedding dimension. The rate of change of distance is defined as the multiple by which the new spatial distance increases relative to the spatial distance at the current embedding dimension.

[0033] The distance change rate is compared with a pre-set judgment value (usually a constant between 10 and 15). If the distance change rate between a target state phase point and its nearest phase point is greater than the judgment value, then the proximity relationship is a projection illusion in low-dimensional space. That is, two phase points that are not adjacent in high-dimensional space exhibit proximity characteristics after being projected into low-dimensional space, which is not the true dynamic proximity relationship of discontinuous cutting.

[0034] The proportion of false neighbors to the total number of neighbors is calculated for each embedding dimension. The embedding dimension at which the proportion of false neighbors first drops to a preset approach threshold is selected as the optimal embedding dimension. The total number of points identified as false neighbors at the current embedding dimension is counted and divided by the total number of points in the phase space state point set to obtain the proportion of false neighbors at the current embedding dimension. This process is repeated for each embedding dimension in an increasing sequence to obtain a sequence of false neighbor proportions for each embedding dimension. The embedding dimension at which the proportion of false neighbors first falls below or equals the preset approach threshold is selected from the sequence of false neighbor proportions and determined as the optimal embedding dimension.

[0035] It should be explained that the approach threshold is set based on a balance between the sufficiency and practicality of phase space reconstruction. Due to the constant noise interference, limited data length, and computational accuracy limitations during intermittent cutting, it is difficult to reduce the proportion of spurious neighbor points to an absolute zero. Therefore, the approach threshold is usually set to 5%, which allows no more than 5% of phase points to still have spurious neighbor phenomena. This ensures that the phase space trajectory is basically fully unfolded, accurately reflecting the nonlinear dynamic characteristics of the system, while avoiding the computational burden and noise amplification caused by choosing an excessively high dimension in pursuit of a perfect zero value.

[0036] Determining the high-dimensional phase space trajectory matrix includes: Using the optimal delay time as the interval step, a number of state point row vectors with the optimal embedding dimension are sequentially extracted from the cutting vibration time series data stream as phase points; all phase points are arranged in time order to obtain a high-dimensional phase space trajectory matrix.

[0037] For example, if the cutting vibration time-series data stream has 5000 vibration amplitudes, the optimal delay time is 3, and the optimal embedding dimension is 4, then the first phase point is composed of the 1st, 4th, 7th, and 10th vibration amplitudes in the cutting vibration time-series data stream; the second phase point is composed of the 2nd, 5th, 8th, and 11th vibration amplitudes; and so on, until the last phase point can take the 5000th vibration amplitude; these extracted state point row vectors are stacked in row order to form a high-dimensional phase space trajectory matrix containing 4991 rows and 4 columns.

[0038] The chaotic feature analysis module is used to iteratively track the phase points in the high-dimensional phase space trajectory matrix to obtain the maximum Lyapunov exponent. Based on the maximum Lyapunov exponent and its changing trend, it identifies whether the current intermittent cutting process is in the pre-chaotic divergence stage and determines the variable impedance control parameters when the pre-chaotic divergence stage is identified.

[0039] Tracking phase points in a high-dimensional phase space trajectory matrix includes: The first phase point in the high-dimensional phase space trajectory matrix is ​​selected as the initial reference phase point. The phase point with the smallest Euclidean distance to the initial reference phase point is then searched within the high-dimensional phase space trajectory matrix and selected as the initial neighboring phase point. The initial reference phase point characterizes the high-dimensional dynamic state of the interrupted cutting process of the superhard cutting tool at the initial observation time, providing a unique and stable initial reference coordinate for the subsequent evolution step size and local divergence rate. The initial neighboring phase points characterize the associated states near the initial observation time of the interrupted cutting of the superhard cutting tool, which have minimal differences from the initial dynamic state. These points form a paired reference group with the initial reference phase point. By synchronously tracking the distance change between the two on the time axis, the local divergence rate in the initial evolution stage can be accurately calculated.

[0040] It should be explained that if multiple phase points have the same Euclidean distance to the initial reference phase point during the search process, and all of them are the minimum value, then the phase point with the earliest time position is selected as the initial neighboring phase point.

[0041] Record the initial proximity distance between the initial reference phase point and the initial neighboring phase point; The initial reference phase point and the initial neighboring phase points are simultaneously shifted backward along the time axis by one evolution step to obtain the evolved reference phase point and the evolved neighboring phase points. Keeping the relative time interval between the initial reference phase point and the initial neighboring phase points in the high-dimensional phase space trajectory matrix unchanged, the initial reference phase point and the initial neighboring phase points are simultaneously shifted backward along the time axis by one evolution step. After the shift, the new phase point corresponding to the row number of the initial reference phase point, incremented by one, is determined as the evolved reference phase point; the new phase point corresponding to the row number of the initial neighboring phase point, incremented by one, is determined as the evolved neighboring phase point.

[0042] It should be explained that the evolution step size is used to characterize the number of units that the reference phase point and the neighboring phase points are pushed back along the time axis each time during the phase space trajectory tracking process; the evolution step size is usually a fixed value of one, that is, one phase point is pushed back each time tracking; this setting can ensure that the trajectory divergence characteristics of each time node are accurately captured when the phase point evolution tracking is performed.

[0043] The evolved neighbor distances are obtained, and the initial and evolved neighbor distances are normalized and logarithmically transformed to obtain the local divergence rate at the current evolution step size. The spatial distance between the evolved reference phase point and its evolved neighboring phase points is calculated using the Euclidean distance algorithm and determined as the evolved neighbor distance. The evolved neighbor distance is divided by the initial neighbor distance to obtain a distance ratio reflecting the degree of trajectory separation; this distance ratio is then transformed using the natural logarithm to form the local divergence rate of the initial reference phase point at the current evolution step size. The local divergence rate characterizes the relative separation rate of adjacent trajectories in the high-dimensional phase space trajectory matrix at a unit evolution step size, and is a precursory microscopic indicator used to determine whether superhard cutting tools tend towards chaos during intermittent cutting.

[0044] The evolved reference phase point is used as the new reference phase point to iteratively track the trajectory matrix in the high-dimensional phase space until the reference phase point reaches the end of the trajectory matrix in the high-dimensional phase space, and the local divergence rate corresponding to each evolution step is determined.

[0045] For example, a high-dimensional phase space trajectory matrix has 4991 phase points. The first tracking reference phase point is in the first row, the second tracking reference phase point is in the second row, the third tracking reference phase point is in the third row, and so on. When the reference phase point reaches the 4991st row, there are no more subsequent phase points in that row, so the tracking process cannot be performed again and terminates.

[0046] The maximum Lyapunov index is obtained by: An evolutionary relationship curve is constructed with the number of evolutionary steps as the x-axis and the local divergence rate as the y-axis; Inflection point detection is performed on the evolutionary relationship curve. The continuous interval between the first and second inflection points is defined as the linear growth region. Starting from the second evolutionary step, the difference between the local divergence rate corresponding to each evolutionary step and the local divergence rate of the previous evolutionary step is calculated sequentially, forming a first-order difference sequence reflecting the rate of change of the local divergence rate with increasing evolutionary steps. Starting from the third evolutionary step, the difference between the first-order difference value corresponding to each evolutionary step and the first-order difference value of the previous evolutionary step is calculated sequentially, forming a second-order difference sequence reflecting the change in the rate of change of the local divergence rate. The first evolutionary step that causes a sign change in the second-order difference value (from positive to negative or from negative to positive) is the location of the first inflection point. The evolutionary steps after the first inflection point continue to search for the next evolutionary step where the sign of the second-order difference value changes, and this evolutionary step is defined as the second inflection point. The number of evolutionary steps and their corresponding local divergence rates between the first and second inflection points are extracted and defined as the linear growth region.

[0047] Before inflection point detection, the evolutionary relationship curve needs to be smoothed to eliminate the interference of local minor fluctuations on inflection point detection. Specifically, for the local divergence rate corresponding to each evolutionary step, the local divergence rate values ​​within a range of several evolutionary steps before and after it are arithmetically averaged, and the average value is taken as the smoothed local divergence rate for that evolutionary step.

[0048] A fitting analysis was performed on the number of evolutionary steps and the corresponding local divergence rate within the linear growth region. The slope of the straight line obtained from the fitting analysis was determined as the maximum Lyapunov exponent. All evolutionary steps within the linear growth region were extracted as the set of independent variables, and the corresponding local divergence rate was extracted as the set of dependent variables. Linear regression analysis was performed on the independent and dependent variable sets using the least squares method to construct the best-fit straight line equation. The slope parameter in the best-fit straight line equation was calculated, and the value of the slope parameter was determined as the maximum Lyapunov exponent characterizing the degree of chaos in the system.

[0049] It should be explained that the larger the maximum Lyapunov exponent value, the faster the divergence of the cutting vibration trajectory, the higher the degree of chaos, and the greater the risk of chipping of the superhard cutting tool; when the maximum Lyapunov exponent is negative, it is in a stable state, and when the exponent approaches zero, the system is close to the chaotic critical state.

[0050] Identify whether the current intermittent cutting process is in the early stages of chaotic divergence, including: The maximum Lyapunov exponents within the sliding analysis window are arranged chronologically to form a historical evolution sequence. The slope of the linear fit in the historical evolution sequence is extracted as the dynamic evolution trend of the intermittent cutting process. Using the time index in the historical evolution sequence as the independent variable and the corresponding maximum Lyapunov exponent value as the dependent variable, the least squares method is used to perform linear regression fitting on the historical evolution sequence. The rate of change of the fitted line is calculated, and the value of the rate of change is determined as the dynamic evolution trend of the current intermittent cutting process.

[0051] It should be explained that the length of the sliding analysis window is determined based on cutting experiment experience, and a balance needs to be struck between trend smoothness and state response delay. For example, the length of the sliding analysis window can be set to store fifty maximum Lyapunov exponents.

[0052] A preset stability critical threshold and a deterioration rate threshold are compared with the maximum Lyapunov exponent and the dynamic evolution trend at the current moment, respectively. If the maximum Lyapunov exponent at the current moment is less than or equal to the stability critical threshold and the dynamic evolution trend is greater than the deterioration rate threshold, then the current discontinuous cutting process is determined to be in the precursor of chaotic divergence.

[0053] Chaotic divergence precursors refer to the critical stage in which the dynamic behavior of superhard cutting inserts transitions from a stable or periodic motion state to a chaotic state during intermittent cutting. This stage occurs before the superhard cutting insert chipps, representing the early critical window period for potential chipping hazards. At this time, the superhard cutting insert has not yet experienced physical chipping damage, but the cutting dynamics state has already shown an irreversible deterioration trend.

[0054] It should be explained that the stability critical threshold is a numerical limit used to determine whether a superhard cutting insert has approached the boundary of a chaotic state during interrupted cutting. The degradation rate threshold is a numerical limit used to determine the rate of degradation of a superhard cutting insert during interrupted cutting. The stability critical threshold and degradation rate threshold can be adjusted differently according to different superhard cutting insert materials, workpiece materials, and cutting parameter ranges. For example, the stability critical threshold and degradation rate threshold can be set to 0.03 and 0.05, respectively.

[0055] In response to the identified precursors of chaotic divergence, an anti-impact parameter set matching the current cutting condition is extracted from a preset control parameter database; the anti-impact parameter set includes a high damping coefficient, a low stiffness coefficient, and a matching inertia coefficient.

[0056] The control parameter database is pre-established for offline experiments and theoretical calculations. For different combinations of cutting conditions (such as different spindle speeds, different depths of cut, different feed rates, different tool teeth, different workpiece materials, etc.), cutting experiments or simulation analyses are conducted to determine the optimal combination of control parameters that can effectively suppress chatter when chaotic divergence precursors appear under the condition.

[0057] For example, the record format stored in the control parameter database is "spindle speed range (3000 rpm to 4000 rpm), cutting depth range (1 to 2 mm), workpiece material type (aluminum alloy)" which corresponds to a set of "high damping coefficient (500 N·s / m), low stiffness coefficient (1000 N / m), matching inertia coefficient (0.05 kg·m²)".

[0058] The servo actuator module is used to respond to variable impedance control parameter commands, adjust the compensation torque of the feed axis in real time, and switch the working mode of the superhard cutting tool through the compensation torque.

[0059] Real-time adjustment of the feed axis compensation torque, including: The theoretical cutting trajectory position of the superhard insert and the current actual position of the feed axis are obtained during the current interrupted cutting process, and differential comparison is performed to determine the position tracking deviation. The theoretical cutting trajectory position represents the spatial position that the superhard insert should reach at the current moment. The theoretical cutting trajectory position is represented by coordinate values ​​in the machine tool coordinate system. A reverse compensation calculation is performed on the theoretical cutting trajectory position of the superhard insert to convert it into the commanded position that each feed axis should reach. At a unified sampling time, the current actual position is read from the position sensors installed on each feed axis. Wherein, position tracking deviation = commanded position that the feed axis should reach - current actual position of the feed axis.

[0060] For example, the current superhard cutting insert is 100 mm long, and the spindle has a vertical structure. The machine tool coordinate system zero point is located at the Z-axis zero point when the spindle end face is at its highest position, and the X-axis and Y-axis zero points are located at the center of the worktable. The theoretical cutting trajectory positions of the superhard insert are read as follows: x-axis (120.35 mm); y-axis (80.42 mm); z-axis (50.00 mm). Based on the theoretical cutting trajectory positions, the commanded positions that each feed axis should reach are calculated as follows: feed axis x (120.35 mm); feed axis y (80.4 mm); feed axis z (-50.00 mm). The current actual positions of each feed axis are: feed axis x (120.28 mm); feed axis y (80.36 mm); feed axis z (-49.96 mm).

[0061] Differential comparison revealed that the tracking deviation of the feed axis x position was +0.05 mm; the tracking deviation of the feed axis y position was +0.02 mm; and the tracking deviation of the feed axis z position was +0.02 mm.

[0062] Based on the change in position tracking deviation over time, the current velocity tracking error and acceleration tracking error of the feed axis are determined. The position tracking deviation of the feed axis is extracted from multiple consecutive sampling moments. For each feed axis, the change in position tracking deviation is obtained by subtracting the position tracking deviation from the previous moment's deviation at the current moment. This change is then divided by the sampling period to obtain the velocity tracking error of the feed axis at the current sampling moment. Similarly, for each feed axis, the change in velocity tracking error is obtained by subtracting the velocity tracking error from the previous moment's velocity tracking error at the current moment. This change is then divided by the sampling period to obtain the acceleration tracking error of the feed axis at the current sampling moment.

[0063] The position tracking error, velocity tracking error, and acceleration tracking error are adjusted and compensated using an anti-impact parameter set to obtain a compensation torque comprising inertia compensation, damping compensation, and stiffness compensation components. The high damping coefficient, low stiffness coefficient, and matching inertia coefficient are extracted from the anti-impact parameter set. The high damping coefficient is multiplied by the current velocity tracking error to obtain the damping compensation component; the low stiffness coefficient is multiplied by the current position tracking error to obtain the stiffness compensation component; and the matching inertia coefficient is multiplied by the current acceleration tracking error to obtain the inertia compensation component. The stiffness compensation component, damping compensation component, and inertia compensation component are algebraically added to obtain the total compensation torque required to be applied to the feed axis at the current sampling moment.

[0064] Switching between operating modes of superhard cutting tools is accomplished through torque compensation, including: The compensation torque is converted into a current command and injected into the current control loop of the servo driver. Based on the current command, the servo driver controls the feed axis to generate an electromagnetic torque corresponding to the compensation torque. The compensation torque is processed by amplitude conversion according to the signal standard adapted by the servo driver to form the current command of the servo driver's current control loop. Based on the received current command, the current control loop of the servo driver controls the inverter to output the corresponding current to the motor stator winding through pulse width modulation technology. The current in the winding interacts with the magnetic field of the motor's permanent magnet, generating an electromagnetic torque on the motor rotor. The electromagnetic torque drives the feed axis to move.

[0065] The electromagnetic torque drives a minute displacement of the feed axis during the intermittent cutting impact of the superhard insert. This minute displacement switches the contact stiffness between the superhard insert and the workpiece from the high-rigidity position-holding mode under normal feed axis operation to a high-damping, flexible buffer mode. At the moment the superhard insert enters the workpiece, the cutting impact force is transmitted to the feed axis through the superhard insert and the spindle. The electromagnetic torque generated by the servo motor interacts with the impact force, causing a slight yielding displacement of the feed axis at the moment of impact. After the impact, the feed axis immediately returns to the high-rigidity position-holding mode, ensuring machining accuracy during the non-impact phase.

[0066] Under normal operating conditions, the feed axis is in a high-rigidity position-holding mode, meaning that regardless of fluctuations in the cutting force, the feed axis strives to maintain the commanded position. In this mode, the superhard insert and workpiece exhibit "hard contact" characteristics, i.e., the high-rigidity position-holding mode. When the feed axis produces a slight retraction displacement, the superhard insert no longer forcibly maintains the theoretical trajectory position but instead yields slightly to the cutting force. This causes the contact stiffness between the superhard insert and workpiece to decrease, exhibiting "flexible contact" characteristics, i.e., the high-damping flexible buffer mode.

[0067] In this embodiment, high-dimensional phase space mapping of cutting vibration data is completed by relying on the phase space reconstruction module; by using the optimal delay time, the information decorrelation scale of discontinuous cutting in the nonlinear evolution process is accurately captured, thereby avoiding phase space trajectory distortion and information redundancy; by determining the optimal embedding dimension, the cutting dynamic trajectory is completely restored with the minimum dimension, so that the high-dimensional phase space trajectory matrix can truly reflect the nonlinear characteristics of discontinuous cutting, thus providing an accurate data base for chaotic feature analysis.

[0068] By iteratively tracking phase points in the high-dimensional phase space trajectory matrix through the chaotic feature analysis module, and constructing evolution relationship curves and extracting the slope of the linear growth region as the maximum Lyapunov exponent, the sensitivity of the intermittent cutting process to initial conditions is quantified. By fitting to obtain the dynamic evolution trend, and combining the stability critical threshold and the deterioration rate threshold for dual-condition judgment, it is possible to identify the critical precursors that are accelerating towards the chaotic state in advance, before the vibration amplitude shows obvious abnormalities and before the chipping actually occurs.

[0069] The servo execution module generates a compensating torque in response to the detected precursors of chaotic divergence, driving the feed axis to produce a slight displacement at the moment of intermittent cutting impact. At the moment of impact, vibration energy is dissipated by increasing damping and reducing stiffness, avoiding stress concentration caused by hard collision between the superhard insert and the workpiece, thus effectively suppressing chipping. At the same time, by using only a slight displacement without tool retraction, the problem of low machining efficiency caused by frequent deceleration or tool retraction is avoided. This achieves the active absorption of intermittent cutting impact energy without interrupting the cutting process, ultimately achieving a dual improvement in machining efficiency and tool life.

[0070] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed in this invention can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0071] In the several embodiments provided by this invention, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only one method, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0072] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

[0073] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An adaptive control system for intermittent cutting to suppress chipping of superhard cutting inserts, characterized in that, include: The signal acquisition module is used to acquire and process the original high-frequency vibration signal during the current intermittent cutting process of the superhard cutting tool to obtain the cutting vibration time sequence data stream. The phase space reconstruction module is used to perform autocorrelation analysis on the cutting vibration time series data stream to determine the optimal delay time, and to determine the optimal embedding dimension through false nearest neighbor detection. Based on the optimal delay time and the optimal embedding dimension, the phase space of the cutting vibration time series data stream is reconstructed to determine the high-dimensional phase space trajectory matrix. The chaotic feature analysis module is used to iteratively track the phase points in the high-dimensional phase space trajectory matrix to obtain the maximum Lyapunov exponent. Based on the maximum Lyapunov exponent and its changing trend, it identifies whether the current intermittent cutting process is in the pre-chaotic divergence stage and determines the variable impedance control parameters when the pre-chaotic divergence stage is identified. The servo actuator module is used to respond to variable impedance control parameter commands, adjust the compensation torque of the feed axis in real time, and switch the working mode of the superhard cutting tool through the compensation torque.

2. The intermittent cutting adaptive control system for suppressing chipping of superhard cutting inserts according to claim 1, characterized in that, The obtained cutting vibration time-series data stream includes: The cutting vibration of the superhard cutting tool during the current intermittent cutting process is acquired in real time, and the cutting vibration is converted into a raw high-frequency vibration signal that is proportional to the vibration acceleration based on the piezoelectric effect. The original high-frequency vibration signal is preprocessed to form a discrete digital signal sequence; the discrete digital signal sequence is arranged in time order to form a cutting vibration time-series data stream.

3. The intermittent cutting adaptive control system for suppressing chipping of superhard cutting inserts according to claim 2, characterized in that, The process of performing autocorrelation analysis to determine the optimal delay time includes: Set several candidate delay times, shift the cutting vibration time series data stream backward according to the step size of each candidate delay time and truncate it by the same length to obtain the delayed truncation sequence and the original truncation sequence under the corresponding candidate delay time; The interdependence between the original truncation sequence and the delayed truncation sequence under each candidate delay time is analyzed to determine the mutual information overlap under different candidate delay times. The mutual information overlap is then connected sequentially in ascending order of candidate delay time to form a mutual information overlap curve. The candidate delay time corresponding to the first time when the mutual information overlap reaches the valley point is selected from the mutual information overlap curve and determined as the optimal delay time.

4. The intermittent cutting adaptive control system for suppressing chipping of superhard cutting inserts according to claim 3, characterized in that, The method for determining the optimal embedding dimension includes: Set an increasing sequence of embedding dimensions, select one embedding dimension from the increasing sequence in turn, and reconstruct the cutting vibration time series data stream in combination with the optimal delay time to obtain the phase space state point set under each embedding dimension; False neighboring points are identified based on the rate of change of distance between each state point and its neighboring points in the state point set of each phase space as the dimension increases, and the proportion of false neighboring points to the number of neighboring points under each embedding dimension is counted. The embedding dimension corresponding to the first drop of the proportion of false neighboring points to a preset approach threshold is selected as the optimal embedding dimension.

5. The intermittent cutting adaptive control system for suppressing chipping of superhard cutting inserts according to claim 4, characterized in that, The determination of the high-dimensional phase space trajectory matrix includes: Using the optimal delay time as the interval step, a number of state point row vectors with the optimal embedding dimension are sequentially extracted from the cutting vibration time series data stream as phase points; all phase points are arranged in time order to obtain a high-dimensional phase space trajectory matrix.

6. The intermittent cutting adaptive control system for suppressing chipping of superhard cutting inserts according to claim 5, characterized in that, The tracking of phase points in the high-dimensional phase space trajectory matrix includes: The first phase point in the high-dimensional phase space trajectory matrix is ​​selected as the initial reference phase point, and the phase point with the smallest Euclidean distance to the initial reference phase point is searched in the high-dimensional phase space trajectory matrix as the initial neighboring phase point; the initial neighboring distance between the initial reference phase point and the initial neighboring phase point is recorded. The initial reference phase point and the initial neighboring phase point are simultaneously shifted backward along the time axis by one evolution step to obtain the evolved reference phase point and the evolved neighboring phase point; Obtain the evolved neighbor distance, normalize and logarithmically transform the initial neighbor distance and the evolved neighbor distance to obtain the local divergence rate at the current evolution step size; The evolved reference phase point is used as the new reference phase point to iteratively track the trajectory matrix in the high-dimensional phase space until the reference phase point reaches the end of the trajectory matrix in the high-dimensional phase space, and the local divergence rate corresponding to each evolution step is determined.

7. The intermittent cutting adaptive control system for suppressing chipping of superhard cutting inserts according to claim 6, characterized in that, The process of obtaining the maximum Lyapunov index includes: An evolutionary relationship curve is constructed with the number of evolutionary steps as the x-axis and the local divergence rate as the y-axis. Inflection point detection is performed on the evolutionary relationship curve, and the continuous interval between the first inflection point and the second inflection point in the evolutionary relationship curve is determined as the linear growth region. The number of evolutionary steps and the corresponding local divergence rate within the linear growth region are fitted and analyzed. The slope of the straight line obtained from the fitting analysis is determined as the maximum Lyapunov exponent.

8. The intermittent cutting adaptive control system for suppressing chipping of superhard cutting inserts according to claim 7, characterized in that, The identification of whether the current intermittent cutting process is in the precursor of chaotic divergence includes: The maximum Lyapunov exponents within the sliding analysis window are arranged in chronological order to form a historical evolution sequence; the linear fitting slope in the historical evolution sequence is extracted as the dynamic evolution trend of the discontinuous cutting process. The preset stability critical threshold and deterioration rate threshold are compared with the maximum Lyapunov exponent and dynamic evolution trend at the current moment, respectively. If the maximum Lyapunov exponent at the current moment is less than or equal to the stability critical threshold and the dynamic evolution trend is greater than the deterioration rate threshold, then the current discontinuous cutting process is determined to be in the precursor of chaotic divergence. In response to the identified precursors of chaotic divergence, an anti-impact parameter set matching the current cutting condition is extracted from a preset control parameter database; the anti-impact parameter set includes a high damping coefficient, a low stiffness coefficient, and a matching inertia coefficient.

9. The intermittent cutting adaptive control system for suppressing chipping of superhard cutting inserts according to claim 8, characterized in that, The real-time adjustment of the feed axis compensation torque includes: The theoretical cutting trajectory position of the superhard insert and the current actual position of the feed axis are obtained in the current interrupted cutting process, and the position tracking deviation is determined by differential comparison; based on the change of position tracking deviation over time, the current speed tracking error and acceleration tracking error of the feed axis are determined. By adjusting and compensating for position tracking deviation, velocity tracking error and acceleration tracking error using an anti-impact parameter set, a compensation torque containing inertia compensation component, damping compensation component and stiffness compensation component is obtained.

10. The intermittent cutting adaptive control system for suppressing chipping of superhard cutting inserts according to claim 9, characterized in that, The switching of the superhard cutting tool's operating mode via compensation torque includes: The compensation torque is converted into a current command and injected into the current control loop of the servo driver; the servo driver controls the feed axis to generate an electromagnetic torque corresponding to the compensation torque based on the current command. The electromagnetic torque drives the feed axis to generate a small displacement during the intermittent cutting impact of the superhard insert. Through this small displacement, the contact stiffness between the superhard insert and the workpiece is switched from the high-rigidity position holding mode under normal feed axis operation to the high-damping flexible buffer mode.

Images