Adaptive intelligent control method for machine tool functional components

Abstract

The application discloses a machine tool functional component adaptive intelligent control method, and belongs to the technical field of intelligent control of numerical control machine tools. The method comprises the following steps: collecting multi-dimensional dynamic response characteristic signals of a machine tool functional component in a machining process, including spindle vibration acceleration sequence, tool holder torque fluctuation sequence and feed shaft displacement deviation sequence; performing time-frequency domain joint feature extraction processing on the multi-dimensional dynamic response characteristic signals to generate a component running state feature set composed of transient impact feature sequence and steady cycle feature sequence; constructing a multi-controller collaborative architecture according to the component running state feature set, including a proportional-integral-derivative control branch and a sliding mode variable structure control branch; calculating the dynamic participation weight of the two branches by using an adaptive switching mechanism to generate a fusion control instruction; and decomposing the fusion control instruction into a drive control component, a brake control component and a damping adjustment component to generate a drive signal sequence of a machine tool functional component execution terminal.

Description

Technical Field

[0001] This invention relates to the field of intelligent control technology for CNC machine tools, specifically to an adaptive intelligent control method for machine tool functional components. Background Technology

[0002] During high-speed, high-precision machining, machine tool components are affected by complex factors such as sudden changes in cutting loads, uneven workpiece material, and nonlinear friction of the servo system. The operating state of these components exhibits a complex interplay of transient impacts and steady-state cycles. Traditional control methods for machine tool components typically employ a single fixed-parameter controller, such as a proportional-integral-derivative (PID) controller that regulates spindle speed, tool turret torque, and feed axis position throughout the machining process. While the PID controller can maintain reasonable control accuracy in the steady-state phase when machining conditions are stable and the system's dynamic characteristics are approximately linear, it becomes vulnerable to large transient shocks caused by sudden changes in cutting force or abrupt tool turret stops and reversals. Due to its limited dynamic response bandwidth, the fixed-parameter PID controller struggles to suppress severe oscillations in a short time, causing the actual motion trajectory of the component to deviate from the commanded value, leading to surface chattering or dimensional deviations. To address these transient shocks, some technical solutions introduce discontinuous control structures, such as sliding mode variable structure control, which forces the system state to converge rapidly along a preset sliding surface using a high-gain switching term. However, the inherent chattering characteristics of sliding mode variable structure control continuously inject high-frequency fluctuations into the actuator during steady-state operation, causing micro-amplitude high-frequency vibrations in the tool holder and feed axis, exacerbating wear on the mechanical transmission chain and reducing surface finish. A single control law cannot simultaneously resolve the contradiction between the two requirements of rapid suppression of transient impacts and smooth steady-state operation. Furthermore, the control output of traditional control architectures typically drives the actuator directly as a single analog or digital signal, without functional decoupling of the control commands. When the control commands contain multiple intentions such as driving, braking, and damping adjustment, the actuator cannot differentiate responses to different control components, leading to concurrent problems such as delayed start-up impulse, overshoot during braking, and insufficient suppression of high-frequency oscillations, thus restricting the dynamic following ability and motion smoothness of functional components in complex trajectory machining. Therefore, the control problems that need to be solved include: how to automatically schedule control laws with different characteristics based on the real-time operating state characteristics of components under varying processing conditions, so as to balance the control performance during transient impact and steady-state operation; and how to accurately decompose the comprehensive control commands according to the driving, braking and damping functions, so that the execution terminal obtains component signals that match the current motion requirements, thereby improving the motion accuracy and stability of components throughout the transient transition and steady-state maintenance process. Summary of the Invention

[0003] This invention provides an adaptive intelligent control method for machine tool functional components. Based on transient impact and steady-state cyclic characteristics extracted from multi-dimensional dynamic response feature signals, it constructs a multi-controller architecture that coordinates proportional-integral-derivative (PID) control branches and sliding mode variable structure (SMS) control branches. An adaptive switching mechanism dynamically adjusts the participation weights of the two branches, enabling the control output to simultaneously possess strong transient robustness suppression and high steady-state precision maintenance capabilities. Furthermore, the fused control command is decomposed in real-time into drive control components, braking control components, and damping adjustment components, generating a composite drive signal sequence required by the execution terminal. This achieves adaptive, precise, and stable control of functional components across all operating conditions during complex machining processes.

[0004] To achieve the above objectives, the present invention provides the following technical solution: The present invention relates to an adaptive intelligent control method for machine tool functional components, comprising: The machine tool's functional components are collected as multidimensional dynamic response characteristic signals during the machining process. These multidimensional dynamic response characteristic signals include spindle vibration acceleration sequence, tool holder torque fluctuation sequence, and feed axis displacement deviation sequence.

[0005] As a technical solution of the present invention, the process of acquiring multi-dimensional dynamic response characteristic signals of machine tool functional components during machining is specifically as follows: A piezoelectric vibration sensor array is set on the surface of the machine tool spindle end cover and tool post base to acquire the spindle vibration acceleration sequence; a Hall current sensor is connected in series on the power supply line of the machine tool spindle drive motor to acquire the time-domain waveform data of the motor phase current; the time-domain waveform data is processed by three-phase current synthesis to generate a tool post torque fluctuation sequence; a high-speed counter is connected to the output end of the grating ruler reading head of the machine tool feed axis to record the actual position feedback value in each interpolation cycle; the difference between the actual position feedback value and the command position value sent by the CNC system is calculated to generate a feed axis displacement deviation sequence; the spindle vibration acceleration sequence, tool post torque fluctuation sequence, and feed axis displacement deviation sequence are synchronized according to the sampling timestamp to generate a time-synchronized multi-dimensional dynamic response characteristic signal. Preferably, when synchronizing and aligning the spindle vibration acceleration sequence, tool holder torque fluctuation sequence, and feed axis displacement deviation sequence according to the sampling timestamp, the sampling clock of the tool holder torque fluctuation sequence is used as the reference clock. Linear interpolation resampling is then performed on the spindle vibration acceleration sequence and the feed axis displacement deviation sequence respectively, so that the three sets of sequences have the same sampling frequency and time coordinate origin. This eliminates the time base deviation between multi-source heterogeneous sensors and improves the consistency and accuracy of subsequent feature extraction.

[0006] A joint time-frequency domain feature extraction process is performed on the multi-dimensional dynamic response characteristic signal to generate a component operating state feature set consisting of transient impact feature sequences and steady-state cycle feature sequences. Specifically, the process involves: performing a short-time Fourier transform on the spindle vibration acceleration sequence to obtain the time-frequency spectrum of the vibration signal; extracting the frequency components corresponding to the energy peaks from the time-frequency spectrum as the spindle transient impact feature sequence; performing wavelet packet decomposition on the tool holder torque fluctuation sequence; arranging the energy coefficients in each frequency band after decomposition in ascending order of frequency to generate a multi-scale energy distribution vector; extracting the frequency band coefficients above a preset energy threshold from the multi-scale energy distribution vector as the torque transient impact feature sequence; performing autocorrelation function calculation on the feed axis displacement deviation sequence; extracting the main lobe width and side lobe attenuation rate of the autocorrelation function; determining the steady-state cycle feature sequence of the feed system based on the main lobe width and side lobe attenuation rate; and concatenating the spindle transient impact feature sequence, torque transient impact feature sequence, and steady-state cycle feature sequence according to feature dimensions to generate the component operating state feature set. Preferably, the wavelet packet decomposition process has 5 decomposition layers, uses the Daubechies wavelet basis function, and sets the preset energy threshold to 1.5 times the average energy of each frequency band. This can effectively suppress background noise interference while preserving the integrity of transient impact characteristics and improve the detection sensitivity of torque anomaly events.

[0007] A multi-controller collaborative architecture is constructed based on the component operating state feature set. This architecture includes a proportional-integral-derivative (PID) control branch and a sliding mode variable structure (SMD) control branch. The construction process is as follows: Transient impact feature sequences are extracted from the component operating state feature set. The initial value of the switching gain of the SMD control branch is determined based on the amplitude change rate of the transient impact feature sequences. Sliding surface parameters are set based on the initial switching gain value, and the reaching law function of the SMD control branch is constructed. Steady-state cyclic feature sequences are extracted from the component operating state feature set. The proportional coefficient, integral time constant, and derivative time constant of the PID control branch are determined based on the periodic stationarity of the steady-state cyclic feature sequences, and the transfer function of the PID control branch is constructed. A parallel connection relationship is established between the output terminals of the SMD control branch and the PID control branch. A weighted fusion node is set at the intersection of the parallel connections. The multi-controller collaborative architecture is constructed based on the input-output mapping relationship of the weighted fusion node.

[0008] In a preferred embodiment, the process of establishing a parallel connection between the output of the sliding mode variable structure control branch and the output of the proportional-integral-derivative (PID) control branch, and setting a weighted fusion node at the intersection point, is as follows: The output of the sliding mode variable structure control branch is connected to the first input of the weighted fusion node via a first signal conditioning circuit, the first signal conditioning circuit including a first limiter and a first impedance matching network; the output of the PID control branch is connected to the second input of the weighted fusion node via a second signal conditioning circuit, the second signal conditioning circuit including a second limiter and a second impedance matching network; a first multiplier, a second multiplier, and an adder are set inside the weighted fusion node to convert the output of the sliding mode variable structure control branch into a parallel connection between the output of the sliding mode variable structure control branch and the output of the proportional-integral-derivative (PID) control branch. The sliding mode variable structure control branch output by the adaptive switching mechanism is weighted and input to the first multiplier input of the first multiplier. The output value of the sliding mode variable structure control branch, processed by the first signal conditioning circuit, is input to the second multiplier input of the first multiplier. The proportional-integral-derivative (PID) control branch output by the adaptive switching mechanism is weighted and input to the first multiplier input of the second multiplier. The output value of the PID control branch, processed by the second signal conditioning circuit, is input to the second multiplier input of the second multiplier. The product outputs of the first and second multipliers are connected to the two adder inputs of the adder, respectively, and the fusion control command is obtained at the output of the adder. Through the limiting and impedance matching processing of the signal conditioning circuit, the level compatibility and impedance matching of the output signals of different control branches at the fusion node are ensured, avoiding the adverse effects of signal distortion on control quality.

[0009] By utilizing the adaptive switching mechanism in a multi-controller collaborative architecture, dynamic participation weight calculation is performed on the proportional-integral-derivative (PID) control branch and the sliding mode variable structure (SMD) control branch to generate fused control commands. The specific generation process is as follows: The component operating state feature set at the current sampling time is obtained; the state feature deviation is calculated based on this feature set, where the deviation is the absolute value of the difference between the transient impact characteristic amplitude at the current time and the transient impact characteristic amplitude at the previous time; based on the comparison between the state feature deviation and a preset deviation threshold, the operating mode of the adaptive switching mechanism is determined, including transient-dominant mode, steady-state-dominant mode, and transitional equilibrium mode; when the adaptive switching mechanism operates in transient-dominant mode, the participation weight of the sliding mode variable structure (SMD) control branch is set to the first weight value, and the participation weight of the PID control branch is set to the second weight value, with the first weight value being greater than the second weight value to leverage the rapid suppression capability of sliding mode control against impulsive disturbances; when the adaptive switching mechanism operates in steady-state-dominant mode, the participation weight of the sliding mode variable structure (SMD) control branch is set to the second weight value, where the first weight value is greater than the second weight value to leverage the rapid suppression capability of sliding mode control against impulsive disturbances; when the adaptive switching mechanism operates in steady-state-dominant mode, the participation weight of the sliding mode variable structure (SMD) control branch is set to the third weight value, and the participation weight of the SMD control branch is set to the fourth weight value, and the participation weight of the SMD control branch is set to the fifth weight value, and the participation weight of the SMD control branch is set to the sixth ... In the state-dominant mode, the participation weight of the proportional-integral-derivative (PID) control branch is set to the third weight value, and the participation weight of the sliding mode variable structure (SMD) control branch is set to the fourth weight value. The third weight value is greater than the fourth weight value to utilize the precise elimination capability of PID control for periodic steady-state errors. When the adaptive switching mechanism operates in transition equilibrium mode, the participation weights of the PID and SMD control branches are linearly adjusted according to the ratio of the difference between the state characteristic deviation and the preset deviation threshold, keeping the sum of the participation weights of the two branches constant. This ensures smooth and continuous control output during the transition process and avoids control quantity jumps during mode switching. Based on the output values ​​of the PID and SMD control branches and the corresponding participation weights of the two branches, a weighted summation calculation is performed to generate a fused control command. Preferably, the state characteristic deviation is the ratio of the absolute value of the difference between the transient impact characteristic amplitude at the current moment and the transient impact characteristic amplitude at the previous moment to the transient impact characteristic amplitude at the previous moment, and the preset deviation threshold is set to 0.15. The deviation calculation method using ratios can adapt to the differences in the base value of characteristic amplitude under different working conditions, making the switching criterion independent of working conditions and improving the generalization ability of the deviation threshold to different processing parameters.

[0010] The fusion control command is decomposed into drive control component, braking control component and damping adjustment component, and the drive signal sequence of the machine tool functional component execution terminal is generated based on the drive control component, braking control component and damping adjustment component. The decomposition and generation process is as follows: The fusion control command is input to a three-channel decomposer, which includes a drive channel, a braking channel, and a damping channel. In the drive channel, the fusion control command is subjected to positive amplitude holding processing, and the signal segments with positive amplitude in the fusion control command are extracted as drive preparation signals. The drive preparation signals are then amplified to generate drive control components. In the braking channel, the fusion control command is subjected to negative amplitude inversion processing, and the signal segments with negative amplitude in the fusion control command are extracted and their absolute values ​​are converted to generate braking preparation signals. The braking preparation signals are then amplified to generate braking control components. In the damping channel, the fusion control command is subjected to high-frequency component extraction processing. Low-frequency components in the fusion control command are filtered out using a high-pass filter, and the high-frequency oscillation components in the fusion control command are extracted as damping preparation signals. The damping preparation signals are then scaled to generate damping adjustment components. The drive control components, braking control components, and damping adjustment components are aligned according to their timing sequence and then distributed to the drive input port, braking input port, and damping adjustment input port of the machine tool functional component execution terminal, respectively, to generate a drive signal sequence. By decoupling a fusion control command into three physically defined execution components, the driving force, braking force, and damping characteristics of the execution terminal can be precisely controlled simultaneously. This allows the machine tool's functional components to obtain force characteristics that match the current working conditions during each stage of acceleration, deceleration, and speed maintenance, thus suppressing oscillations and overshoot during motion.

[0011] This invention acquires multi-dimensional dynamic response feature signals including vibration, torque, and displacement deviation, and uses joint time-frequency domain feature extraction to obtain a feature set of component operating states reflecting transient impact and steady-state cycle characteristics. Based on this, a multi-controller collaborative architecture is constructed with a parallel proportional-integral-derivative (PID) control branch and a sliding mode variable structure (SMD) control branch. The participation weights of the two control branches are adaptively adjusted according to changes in the machining state to generate fused control commands. Finally, the fused control commands are decomposed into three execution components: drive, braking, and damping. This achieves multi-mode adaptive control of machine tool functional components. When the cutting force changes abruptly, the sliding mode control quickly intervenes to suppress transient impacts. During the stable cutting stage, the PID control continuously compensates for steady-state deviations. During the transition stage, the control continuity is ensured through smooth weight switching. This significantly improves the response speed, stability, and anti-disturbance capability of machine tool functional components under complex and variable working conditions.

[0012] The technical effects and advantages provided by the present invention in the above technical solution are as follows: Transient impact characteristic sequences and steady-state cyclic characteristic sequences are extracted from the spindle vibration acceleration sequence, tool holder torque fluctuation sequence, and feed axis displacement deviation sequence through joint time-frequency domain feature extraction. The operating states of the components are then finely divided according to their frequency domain structure and time-varying characteristics. Based on the amplitude change rate of the transient impact characteristic sequence, the switching gain and reaching law function of the sliding mode variable structure control branch are determined, enabling the strength of the sliding mode control law to adaptively adjust with the severity of the impact. When an impact occurs, a high-gain correction effect can be applied quickly, significantly shortening the transient recovery time. Based on the periodic stationary characteristics of the steady-state cyclic characteristic sequence, the proportional coefficient, integral time constant, and derivative time constant of the proportional-integral-derivative control branch are determined, allowing the control branch to eliminate steady-state residual deviations and maintain motion stability during the steady-state operation phase through optimized linear adjustment. The adaptive switching mechanism introduced in the multi-controller collaborative architecture dynamically allocates the participation weights of the two branches based on the deviation of state characteristics. In transient dominant mode, the weight of the sliding mode variable structure control branch is increased, while the weight of the proportional-integral-micro branch is decreased, ensuring that shock disturbances are quickly and forcibly suppressed. In steady-state dominant mode, the weight of the proportional-integral-micro branch is increased, while the weight of the sliding mode variable structure branch is decreased, avoiding the transmission of sliding mode chattering to the actuator. In transitional equilibrium mode, the weights are continuously and linearly adjusted according to the deviation, ensuring that there are no abrupt changes in control law switching. It takes into account the two contradictory performance indicators of strong transient robustness and high steady-state accuracy in the whole domain, eliminating the blind spots of a single controller under complex working conditions. The fused control commands are functionally decoupled into drive control, braking control, and damping adjustment components by a three-channel decomposer. The drive channel, through positive amplitude holding and power amplification, extracts only the positive drive component and enhances its energy, providing the actuator with a clear acceleration drive signal and eliminating drive force delay. The braking channel reconstructs braking energy through negative amplitude reversal and absolute value transformation, allowing deceleration and stop commands to be transmitted independently, preventing braking overshoot or trailing caused by crosstalk between braking and drive components. The damping channel extracts high-frequency oscillation components through a high-pass filter and performs amplitude scaling to form a damping injection signal that matches the vibration frequency, actively attenuating the high-frequency residual oscillations generated by the actuator after transient excitation. The three functional components are applied separately to the drive input port, braking input port, and damping adjustment input port, enabling the actuator to respond to the triple demands of acceleration, deceleration, and vibration suppression in parallel within the same control cycle. This overcomes the shortcomings of traditional single control quantities, which cannot simultaneously address power output and vibration suppression, improving the dynamic tracking accuracy and smoothness of functional components in complex machining trajectories. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0014] Figure 1 This is a flowchart of the adaptive intelligent control method for machine tool functional components; Figure 2 This is a flowchart of the acquisition and synchronous processing of multi-dimensional dynamic response characteristic signals of machine tool functional components; Figure 3 This is a flowchart of the process for generating the component operating status feature set; Figure 4 This is a flowchart of a multi-controller collaborative control method based on transient impact and steady-state cycle characteristics; Figure 5 It is a flowchart of the parallel processing of the three channels of drive, braking and damping of the integrated control command. Detailed Implementation

[0015] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, 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.

[0016] See Figure 1 This invention provides an adaptive intelligent control method for machine tool functional components. The method collects multi-dimensional dynamic response characteristic signals of machine tool functional components during the machining process, performs joint time-frequency domain feature extraction to generate a component operating state feature set, and then constructs a multi-controller collaborative architecture including proportional-integral-derivative control branches and sliding mode variable structure control branches. The method uses an adaptive switching mechanism to dynamically calculate the participation weights of the two branches and generate a fused control command. Finally, the fused control command is decomposed into drive control components, braking control components, and damping adjustment components to drive the execution terminal of the machine tool functional components.

[0017] Specifically, the acquired multidimensional dynamic response feature signals include spindle vibration acceleration sequences, tool holder torque fluctuation sequences, and feed axis displacement deviation sequences. Time-frequency domain joint feature extraction is performed on these signals to obtain a component operating state feature set composed of transient impact feature sequences and steady-state cyclic feature sequences. In the multi-controller collaborative architecture built based on this feature set, the proportional-integral-derivative (PID) control branch and the sliding mode variable structure control branch are connected in parallel to the same weighted fusion node. An adaptive switching mechanism calculates the dynamic participation weights of the two branches in real time according to the component operating state, and generates a fused control command after weighted fusion. The fused control command is processed by a three-channel decomposer to obtain drive control components, braking control components, and damping adjustment components. These three components correspond to the drive input port, braking input port, and damping adjustment input port of the machine tool functional component execution terminal, respectively, forming the final drive signal sequence, thereby achieving high-performance adaptive control of the machine tool functional components.

[0018] Example 1: In specific implementation, please refer to Figure 2 The system collects multidimensional dynamic response characteristic signals of machine tool functional components during the machining process. These multidimensional dynamic response characteristic signals include spindle vibration acceleration sequence, tool holder torque fluctuation sequence, and feed axis displacement deviation sequence.

[0019] A piezoelectric vibration sensor array is installed on the surface of the machine tool spindle end cover and tool post base to acquire the spindle vibration acceleration sequence. The mounting plane of each piezoelectric vibration sensor in the array is perpendicular to the spindle axis. The spacing between adjacent piezoelectric vibration sensors is evenly distributed according to the geometry of the spindle end cover and tool post base, ensuring the array covers the critical vibration transmission paths of the spindle tip and tool post mounting surface. The charge signals output from each channel of the piezoelectric vibration sensor array are converted into voltage signals by a charge amplifier and then converted from analog to digital by a synchronous data acquisition module at a preset sampling frequency to generate the spindle vibration acceleration sequence. The preset sampling frequency is set based on the analysis bandwidth of the vibration signal at the spindle's highest speed, and is set to 4 to 5 times the analysis bandwidth of the vibration signal.

[0020] Hall current sensors are connected in series on the power supply line of the machine tool spindle drive motor to collect time-domain waveform data of the motor phase currents. The Hall current sensors are connected in series on the U-phase, V-phase, and W-phase power supply lines of the spindle drive motor, and simultaneously collect the instantaneous values ​​of the U-phase, V-phase, and W-phase currents. When performing three-phase current synthesis processing on the time-domain waveform data, the instantaneous values ​​of the U-phase, V-phase, and W-phase currents are input to a transformation module from a three-phase stationary coordinate system to a two-phase rotating coordinate system. The transformation module performs coordinate transformation based on the current electrical angle of the spindle drive motor and outputs quadrature-axis and direct-axis current components. The quadrature-axis current component is multiplied by the electromagnetic torque coefficient of the spindle drive motor to obtain the instantaneous electromagnetic torque value output by the spindle drive motor. These instantaneous electromagnetic torque values ​​are arranged according to the sampling time to generate a tool post torque fluctuation sequence.

[0021] A high-speed counter is connected to the output of the grating ruler reading head on the machine tool's feed axis. This high-speed counter records the actual position feedback value for each interpolation cycle. The quadrature coded pulse signal output from the grating ruler reading head is connected to the A-phase and B-phase inputs of the high-speed counter. The high-speed counter simultaneously counts the rising and falling edges of the quadrature coded pulse signal in a four-fold frequency multiplication mode to obtain the actual position feedback value for each interpolation cycle. The difference between the actual position feedback value and the command position value sent by the CNC system is calculated. The command position value is the target position value generated by the CNC system through interpolation calculation at the beginning of each interpolation cycle. The difference calculation involves subtracting the command position value from the actual position feedback value. The resulting difference sequence is the feed axis displacement deviation sequence.

[0022] The spindle vibration acceleration sequence, tool holder torque fluctuation sequence, and feed axis displacement deviation sequence are synchronized and aligned according to their sampling timestamps to generate a time-synchronized multidimensional dynamic response characteristic signal. During the synchronization and alignment process, the sampling clock of the tool holder torque fluctuation sequence is used as the reference clock, and linear interpolation resampling is performed on the spindle vibration acceleration sequence and the feed axis displacement deviation sequence respectively, so that the three sets of sequences have the same sampling frequency and time coordinate origin.

[0023] The specific implementation method of using the sampling clock of the tool holder torque fluctuation sequence as the reference clock is as follows: read the timestamp of each sampling point in the tool holder torque fluctuation sequence, and set the time starting point of the tool holder torque fluctuation sequence sampling clock as the unified time coordinate origin. When performing linear interpolation resampling processing on the spindle vibration acceleration sequence, for the sampling time and sampling value of two adjacent original sampling points in the spindle vibration acceleration sequence, according to the target sampling time required by the reference clock of the tool holder torque fluctuation sequence, perform linear interpolation calculation between the two adjacent original sampling points to obtain the resampled value corresponding to the target sampling time. The linear interpolation calculation is performed in the following way: In the sequence of set spindle vibration acceleration, the first... The sampling time of each original sampling point is The sampled value is , No. The sampling time of each original sampling point is The sampled value is The target sampling time is ,and Then the target sampling time corresponding resampled value The calculation formula is: in, This indicates the sequence of spindle vibration accelerations at the target sampling time. The linear interpolated resampled value; The sequence of principal shaft vibration accelerations is represented in the 1st... The sampled values ​​of each original sampling point; The sequence of principal shaft vibration accelerations is represented in the 1st... The sampled values ​​of each original sampling point; Represents the sequence of principal shaft vibration accelerations. The sampling time of each original sampling point; Represents the sequence of principal shaft vibration accelerations. The sampling time of each original sampling point; This indicates the target sampling time required by the reference clock for the tool holder torque fluctuation sequence.

[0024] When performing linear interpolation resampling on the feed axis displacement deviation sequence, the same calculation method as the spindle vibration acceleration sequence is used. The sampling time and deviation value of adjacent original sampling points in the feed axis displacement deviation sequence are substituted into the above linear interpolation calculation formula to obtain a feed axis displacement deviation resampling sequence with the same sampling frequency and time coordinate origin as the tool holder torque fluctuation sequence.

[0025] After the three sets of sequences are synchronized and aligned, a time-synchronized multidimensional dynamic response feature signal is generated. The spindle vibration acceleration sequence, tool holder torque fluctuation sequence, and feed axis displacement deviation sequence in the multidimensional dynamic response feature signal have a one-to-one corresponding timestamp at each sampling moment.

[0026] Example 2: In specific implementation, please refer to Figure 3 The short-time Fourier transform of the spindle vibration acceleration sequence is performed to obtain the time spectrum of the vibration signal. The frequency components corresponding to the energy peaks are extracted from the time spectrum as the transient impact characteristic sequence of the spindle.

[0027] The short-time Fourier transform (SFT) processing is implemented as follows: The window function type is set to a Hanning window, the window function length is set to 256 sampling points, and the window function sliding step size is set to 128 sampling points. The main shaft vibration acceleration sequence is divided into multiple overlapping time frames according to the sliding step size, with 256 sampling points in each time frame. The 256 sampling points in each time frame are multiplied by the Hanning window function and then subjected to a fast Fourier transform (FFT) to obtain the spectrum vector corresponding to each time frame. The spectrum vectors of all time frames are arranged chronologically along the time axis to form a time-spectrum diagram of the vibration signal. The horizontal axis of the time-spectrum diagram represents the time frame number, the vertical axis represents the frequency value, and the grayscale value or color value represents the spectral amplitude.

[0028] The steps for extracting the frequency components corresponding to the energy peaks from the time-spectrum graph as the principal axis transient impact feature sequence are as follows: Within each time frame of the time-spectrum graph, search for the frequency point with the largest spectral amplitude and record its frequency value and spectral amplitude. Arrange the frequency values ​​found within each time frame in time frame order to generate a frequency value sequence; arrange the spectral amplitudes found within each time frame in time frame order to generate a spectral amplitude sequence. The frequency value sequence and the spectral amplitude sequence together constitute the principal axis transient impact feature sequence, where the frequency value sequence represents the trend of the main frequency components changing over time during the transient impact event, and the spectral amplitude sequence represents the trend of the intensity of the transient impact event changing over time.

[0029] Wavelet packet decomposition was performed on the tool holder torque fluctuation sequence, with a decomposition level of 5. The Daubechies wavelet basis function was used to obtain the energy coefficients of each frequency band. The order of the Daubechies wavelet basis function was set to 4, corresponding to a filter length of 8 coefficients. The wavelet packet decomposition process was as follows: the tool holder torque fluctuation sequence was used as the original signal of level 0. In level 1 decomposition, the original signal of level 0 was passed through a low-pass filter and a high-pass filter, and after downsampling, the level 1 low-frequency sub-band signal and high-frequency sub-band signal were obtained. In level 2 decomposition, the level 1 low-frequency sub-band signal was passed through a low-pass filter and a high-pass filter, and after downsampling, the level 2 low-frequency sub-band signal and high-frequency sub-band signal were obtained. The level 1 high-frequency sub-band signal was passed through a low-pass filter and a high-pass filter, and after downsampling, the level 2 low-frequency sub-band signal and high-frequency sub-band signal were obtained. This process was repeated until level 5 decomposition was completed, resulting in 32 frequency band sub-band signals. For each frequency band sub-band signal, calculate the sum of the squares of the amplitudes of all sampling points within that frequency band sub-band signal, and use this sum as the energy coefficient of that frequency band.

[0030] The energy coefficients within each frequency band after decomposition are arranged in ascending order of frequency to generate a multi-scale energy distribution vector. This vector is 32-dimensional; the first element corresponds to the energy coefficient of the lowest frequency band in the fifth layer, and the 32nd element corresponds to the energy coefficient of the highest frequency band in the fifth layer. Frequency band coefficients above a preset energy threshold in the multi-scale energy distribution vector are extracted as a torque transient impact feature sequence. The preset energy threshold is set to 1.5 times the average energy value of each frequency band. The average energy value of each frequency band is calculated as follows: sum the 32 energy coefficients in the multi-scale energy distribution vector to obtain the total energy coefficient; divide the total energy coefficient by 32 to obtain the average energy value of each frequency band; multiply the average energy value of each frequency band by 1.5 to obtain the preset energy threshold value. The relationship between each energy coefficient in the multi-scale energy distribution vector and the preset energy threshold value is compared one by one, and the frequency band coefficients corresponding to energy coefficients greater than the preset energy threshold are marked as extraction targets. The extracted objects are represented by the frequency band index number and the energy coefficient of the frequency band. All extracted objects are arranged in ascending order of frequency band index number, forming a torque transient impact feature sequence.

[0031] The autocorrelation function is calculated on the feed axis displacement deviation sequence to extract the main lobe width and side lobe attenuation rate of the autocorrelation function. The steady-state cyclic characteristic sequence of the feed system is determined based on the main lobe width and side lobe attenuation rate.

[0032] The autocorrelation function calculation is implemented by setting the length of the feed axis displacement deviation sequence as follows: The sequence of feed axis displacement deviations is represented by sampling points as follows: ,in The sampling point number, Calculate the autocorrelation function value of the feed axis displacement deviation sequence. The formula for calculating the autocorrelation function value is: in, Indicates the number of lag steps. The autocorrelation function value at time; This represents the total number of sampling points in the feed axis displacement deviation sequence; Indicates the first position in the feed axis displacement deviation sequence The deviation value of each sampling point; Indicates the first position in the feed axis displacement deviation sequence The deviation value of each sampling point; Indicates the number of lag steps. ; This is an accumulated variable representing the sampling point sequence number.

[0033] The steps for extracting the main lobe width and side lobe attenuation rate of the autocorrelation function are as follows: [The text abruptly ends here, likely due to an incomplete sentence or a formatting error.] By lag steps The autocorrelation function curves are plotted in ascending order of lag number. The maximum value is obtained at the point of lag, with the number of lag steps. As the value increases, the autocorrelation function curve decays from its maximum value towards both sides. From the lag step count... Starting from the initial value, search in the direction of increasing lag steps for the lag function value to first drop to half of its maximum value. Multiply this lag step value by the sampling period and record it as the main lobe width. After finding the lag step value corresponding to the main lobe width, continue searching in the direction of increasing lag steps for the first local peak point in the autocorrelation function curve. Record the autocorrelation function value and the corresponding lag step value of this local peak point. Calculate the logarithm of the ratio of the autocorrelation function value of this local peak point to the maximum value of the autocorrelation function. Divide this logarithm by the difference between the lag step value corresponding to the local peak point and the lag step value corresponding to the main lobe width to obtain the sidelobe attenuation rate.

[0034] The steady-state cyclic characteristic sequence of the feed system is determined by combining the main lobe width and the side lobe attenuation rate to form a two-element eigenvector, which is the steady-state cyclic characteristic sequence. The main lobe width reflects the duration of the fundamental component of the periodic error of the feed system, while the side lobe attenuation rate reflects the relative intensity decay rate of the harmonic components in the periodic error of the feed system.

[0035] The transient impact feature sequence of the main shaft, the transient torque impact feature sequence, and the steady-state cycle feature sequence are concatenated according to feature dimensions to generate a component operating state feature set. The feature dimension concatenation process is as follows: the frequency value sequence, the spectral amplitude sequence, the frequency band index sequence of the extracted object, the energy coefficient sequence of the extracted object, and the main lobe width and side lobe attenuation rate from the steady-state cycle feature sequence are sequentially concatenated horizontally to form a vector-like component operating state feature set.

[0036] Example 3: In specific implementation, please refer to Figure 4 The transient impact feature sequence is extracted from the component operating state feature set. The component operating state feature set is in vector form. The spectral amplitude sequence from the main shaft transient impact feature sequence and the energy coefficient sequence from the torque transient impact feature sequence are extracted from the vector according to a preset dimension index. The spectral amplitude sequence and the energy coefficient sequence are merged to form the transient impact feature sequence. Each sampling time in the transient impact feature sequence corresponds to a transient impact feature amplitude.

[0037] The initial value of the switching gain of the sliding mode variable structure control branch is determined based on the amplitude change rate of the transient impact characteristic sequence. The amplitude change rate is calculated as follows: In the transient impact characteristic sequence, the transient impact characteristic amplitude at the current sampling moment is subtracted from the transient impact characteristic amplitude at the previous sampling moment to obtain the amplitude difference. The absolute value of the amplitude difference is divided by the sampling period to obtain the amplitude change rate at the current sampling moment. The amplitude change rate is obtained by traversing all sampling moments of the transient impact characteristic sequence and selecting the largest amplitude change rate from all sampling moments. The largest amplitude change rate is multiplied by a preset gain mapping coefficient to obtain the initial value of the switching gain of the sliding mode variable structure control branch. The preset gain mapping coefficient is determined based on the ratio of the maximum allowable control gain of the machine tool functional component drive system to the maximum transient impact change rate. The maximum allowable control gain is obtained from the technical specifications of the machine tool functional component drive system. This maximum allowable control gain is divided by the historical maximum transient impact change rate recorded under the test conditions to obtain the value of the preset gain mapping coefficient.

[0038] The sliding surface parameters are set based on the initial value of the switching gain. The sliding surface parameters include a sliding surface coefficient vector, the dimension of which is the same as the dimension of the system state variables. The system state variables are defined as the following error and its first derivative. The sliding surface coefficient vector is a two-dimensional vector, where the first-dimensional coefficients correspond to the following error component, and the second-dimensional coefficients correspond to the first derivative component of the following error. The second-dimensional coefficient is set to 1, and the first-dimensional coefficient is set to the initial value of the switching gain multiplied by a dimension conversion factor between the following error component and the first derivative component of the following error in the system state variables. This dimension conversion factor is determined by the ratio of the maximum drive speed of the machine tool functional component's execution terminal to the maximum following error.

[0039] Construct the reaching law function for the sliding mode variable structure control branch. The reaching law function adopts an exponential reaching law form, and its expression is: in, Represents the sliding surface function The derivative with respect to time; Represents the sliding surface function. It is obtained by performing an inner product operation between the system state variables and the sliding surface coefficient vector; Represents the coefficient of the isotropic approaching term. The value is the initial value of the switching gain multiplied by the first preset scaling factor, which is set to 0.1. It is determined according to the typical parameter selection rules for suppressing chattering in sliding mode control theory, so as to ensure the speed of approaching motion while reducing chattering amplitude. Represents the coefficient of the exponential approach term. The value is the initial value of the switching gain multiplied by the second preset scaling factor, which is set to 5 to accelerate the approach rate when the system state moves away from the sliding surface. Represents a sign function, when When greater than zero The value is 1, when When equal to zero The value is 0, when When less than zero The value is -1.

[0040] Steady-state cyclic feature sequences are extracted from the component operating state feature set. Within the component operating state feature set vector, the main lobe width and side lobe attenuation rate are extracted according to a preset dimension index. The main lobe width and side lobe attenuation rate together constitute the steady-state cyclic feature sequence.

[0041] The proportional coefficient, integral time constant, and derivative time constant of the proportional-integral-derivative (PID) control branch are determined based on the periodic stationary characteristics of the steady-state cyclic characteristic sequence. The main lobe width reflects the duration of the fundamental component of the periodic error in the feed system; a larger main lobe width indicates a longer principal period in the periodic stationary characteristics. The proportional coefficient of the PID control branch is set to the reciprocal of the main lobe width multiplied by the proportional mapping gain. The proportional mapping gain is determined based on the steady-state stiffness coefficient of the machine tool functional component's execution terminal, which is the ratio of the output force to the displacement of the execution terminal, obtained from the machine tool functional component's technical parameter table. The integral time constant of the PID control branch is set to the main lobe width multiplied by the first period coefficient, which is set to 2, to ensure that the integral time constant covers at least twice the principal period length, guaranteeing that the integral action can effectively eliminate periodic steady-state errors. The differential time constant of the proportional-integral-derivative control branch is set to the reciprocal of the sidelobe decay rate multiplied by the second period coefficient, which is 0.5, to match the lead time window of the differential action with the harmonic decay time scale characterized by the sidelobe decay rate.

[0042] Construct the transfer function of the proportional-integral-derivative (PID) control branch. The transfer function of the PID control branch in the continuous domain is the sum of the proportional coefficient multiplied by 1 and the inverse of the integral time constant multiplied by the inverse of the Laplace operator, plus a term for the differential time constant multiplied by the Laplace operator. In practical discretization, the transfer function of the PID control branch is converted into a difference equation using the backward difference transform method.

[0043] Establish a parallel connection between the output of the sliding mode variable structure control branch and the output of the proportional-integral-derivative control branch, and set a weight fusion node at the intersection of the parallel connection.

[0044] In practical implementation, the output of the sliding mode variable structure control branch is connected to the first input of the weighted fusion node via a first signal conditioning circuit. The first signal conditioning circuit includes a first limiter and a first impedance matching network. The upper limit of the first limiter is set to 1.2 times the theoretical maximum output value of the sliding mode variable structure control branch, and the lower limit is set to 1.2 times the theoretical minimum output value. The theoretical maximum and minimum output values ​​of the sliding mode variable structure control branch are determined by the calculation results of the reaching law function under the condition of maximum system state deviation. The input impedance of the first impedance matching network is set to the nominal value of the output impedance of the sliding mode variable structure control branch, and the output impedance of the first impedance matching network is set to the nominal value of the input impedance of the first input of the weighted fusion node, thereby achieving impedance matching for signal transmission.

[0045] The output of the proportional-integral-derivative (PID) control branch is connected to the second input of the weighted fusion node via a second signal conditioning circuit. The second signal conditioning circuit includes a second limiter and a second impedance matching network. The upper limit of the second limiter is set to the output saturation value of the PID control branch, which is determined by the maximum output voltage limit of the drive power supply. The lower limit of the second limiter is set to the negative saturation value of the PID control branch, where the negative saturation value is the opposite of the output saturation value. The input impedance of the second impedance matching network is set to the nominal value of the output impedance of the PID control branch, and the output impedance of the second impedance matching network is set to the nominal value of the input impedance of the second input of the weighted fusion node.

[0046] A first multiplier, a second multiplier, and an adder are configured within the weight fusion node. The sliding mode variable structure control branch weights output from the adaptive switching mechanism are input to the first multiplier input, and the sliding mode variable structure control branch output value (processed by the first signal conditioning circuit) is input to the second multiplier input. Similarly, the proportional-integral-derivative (PID) control branch weights output from the adaptive switching mechanism are input to the first multiplier input, and the PID control branch output value (processed by the second signal conditioning circuit) is input to the second multiplier input. Both the sliding mode variable structure control branch weights and the PID control branch weights output from the adaptive switching mechanism are dimensionless coefficients, with values ​​limited to between 0 and 1.

[0047] The product outputs of the first and second multipliers are connected to the two addend inputs of the adder, respectively. The adder sums the product values ​​output by the first and second multipliers, and obtains the fusion control command at the adder's output. Based on the input-output mapping relationship of the weighted fusion nodes, a multi-controller collaborative architecture is constructed.

[0048] Example 4: In practical implementation, the component operating state feature set at the current sampling moment is obtained. This feature set is generated through joint time-frequency domain feature extraction and feature dimension concatenation. The feature set includes transient impact feature sequences and steady-state cycle feature sequences. The transient impact feature amplitude corresponding to the current sampling moment is extracted from the component operating state feature set, denoted as... Extract the transient impact characteristic amplitude corresponding to the previous sampling time, denoted as... .

[0049] The state feature deviation is calculated based on the component operating state feature set at the current sampling time. The formula for calculating the state feature deviation is as follows: in, Indicates the degree of deviation of state characteristics; This represents the transient impact characteristic amplitude at the current sampling moment; This represents the transient impact characteristic amplitude at the previous sampling time. This represents the absolute value of the difference between the transient impact characteristic amplitude at the current sampling time and the transient impact characteristic amplitude at the previous sampling time. When When the value is 0, if If the value is not 0, then the state characteristic deviation degree will be... Set to 1; if If the value is also 0, then the state characteristic deviation degree will be... Set to 0.

[0050] The preset deviation threshold is set to 0.15. The basis for setting the preset deviation threshold to 0.15 is as follows: According to the statistical analysis of transient impact events during machine tool processing, when the relative change rate of the transient impact characteristic amplitude exceeds 15%, it indicates that the processing state has undergone significant transient changes, requiring the rapid intervention of the sliding mode variable structure control branch; when the relative change rate is less than 15%, it indicates that the processing state tends to be stable, suitable for fine adjustment by the proportional-integral-derivative control branch.

[0051] Deviation of state characteristics The working mode of the adaptive switching mechanism is determined by comparing it with a preset deviation threshold.

[0052] A transition range is set in the neighborhood of a preset deviation threshold. The lower limit of the transition range is set to the preset deviation threshold minus the preset hysteresis width, and the upper limit is set to the preset deviation threshold plus the preset hysteresis width. The preset hysteresis width is set to 0.02. The reason for setting the preset hysteresis width to 0.02 is to avoid frequent switching of the working mode due to small fluctuations in the state characteristic deviation near the preset deviation threshold. A hysteresis width of 0.02 can cover the normal measurement noise and calculation fluctuation range.

[0053] When the state characteristic deviation When the value exceeds the sum of the preset deviation threshold and the preset hysteresis width, i.e. At that time, the adaptive switching mechanism is determined to operate in transient-dominant mode.

[0054] When the state characteristic deviation When the value is less than the difference between the preset deviation threshold and the preset hysteresis width, that is... At that time, the adaptive switching mechanism is determined to operate in a steady-state dominant mode.

[0055] When the state characteristic deviation When it is within the transition range, that is At that time, the adaptive switching mechanism is set to transitional equilibrium mode.

[0056] When the adaptive switching mechanism operates in transient-dominant mode, the participation weight of the sliding mode variable structure control branch is set to the first weight value, which is 0.9; the participation weight of the proportional-integral-derivative control branch is set to the second weight value, which is 0.1. The first weight value is greater than the second weight value, making the output of the sliding mode variable structure control branch dominate the fused control command, so as to quickly suppress transient impacts.

[0057] When the adaptive switching mechanism operates in steady-state dominant mode, the participation weight of the proportional-integral-derivative (PID) control branch is set to the third weight value, which is 0.9; the participation weight of the sliding mode variable structure control branch is set to the fourth weight value, which is 0.1. The third weight value is greater than the fourth weight value, ensuring that the output of the PID control branch dominates the fused control command, thereby eliminating steady-state errors and periodic fluctuations.

[0058] When the adaptive switching mechanism operates in transitional equalization mode, it is based on the deviation of state characteristics. The difference ratio from the preset deviation threshold is used to linearly adjust the participation weights of the proportional-integral-derivative control branch and the sliding mode variable structure control branch, ensuring that the sum of the two participation weights remains constant, with a constant value of 1. The specific calculation method for linear adjustment is as follows: the participation weight of the sliding mode variable structure control branch... Calculated using the following formula: in, This represents the participation weight assigned to the sliding mode variable structure control branch under the transition equilibrium mode; This represents the deviation of the state characteristics calculated at the current sampling time; the constant 0.5 is the base value for the weighting of the sliding mode variable structure control branch at the preset deviation threshold; 0.15 is the value of the preset deviation threshold; 0.02 is the half-width of the transition interval, determined by the preset hysteresis width; 0.4 is the weight adjustment amplitude, obtained by multiplying the difference between the first weight value 0.9 and the base value 0.5 by 2, so that at the upper boundary of the transition interval... Reaching 0.9, at the lower boundary of the transition interval. It reached 0.1.

[0059] Participation weights of proportional-integral-derivative control branches The calculation method is as follows: in, This represents the participation weight assigned to the proportional-integral-derivative control branch in the transition equilibrium mode.

[0060] According to the above calculation method, under the transitional equilibrium mode, the deviation of state characteristics The closer to the upper limit of 0.17, the greater the participation weight of the sliding mode variable structure control branch. The closer the value is to 0.9, the greater the participation weight of the proportional-integral-derivative control branch. The closer to 0.1, the greater the deviation of the state characteristics. The closer to the lower limit of 0.13, the greater the participation weight of the proportional-integral-derivative control branch. The closer the value is to 0.9, the greater the participation weight of the sliding mode variable structure control branch. The closer it is to 0.1; at the preset deviation threshold of 0.15, the weights of both types of participants are 0.5.

[0061] Obtain the output values ​​of the proportional-integral-derivative (PID) control branch and the sliding mode variable structure (SMD) control branch at the current sampling time. The output value of the PID control branch is the voltage signal value output by the second signal conditioning circuit, and the output value of the SMD control branch is the voltage signal value output by the first signal conditioning circuit.

[0062] Based on the output values ​​of the proportional-integral-derivative (PID) control branch, the output values ​​of the sliding mode variable structure (SMD) control branch, and the corresponding participation weights of the two branches, a weighted summation calculation is performed to generate a fused control command. The calculation method for the weighted summation is as follows: in, This indicates the generated fusion control command; This represents the participation weight of the proportional-integral-derivative (PID) control branch, which is determined by the second weight value, the third weight value, or the weight under the transition equilibrium mode, depending on the operating mode. ; This represents the output value of the proportional-integral-derivative (PI-DE) control branch. This indicates the participation weight of the sliding mode variable structure control branch, which is determined by the first weight value, the fourth weight value, or the weight under the transition equilibrium mode, depending on the operating mode. ; This represents the output value of the sliding mode variable structure control branch.

[0063] Through the above weighted summation calculation, the fusion control command at the current sampling time is obtained, and the fusion control command is output to the three-channel decomposer for subsequent decomposition processing.

[0064] Example 5: In specific implementation, please refer to Figure 5 The fusion control command is input to a three-channel resolver, which includes a drive channel, a braking channel, and a damping channel. The three-channel resolver internally has signal branch nodes. The fusion control command enters the drive channel, braking channel, and damping channel simultaneously from the signal branch nodes, and the three channels process the fusion control command in parallel.

[0065] In the drive channel, positive amplitude hold processing is performed on the fusion control command. The positive amplitude hold processing works as follows: at each sampling moment, the instantaneous amplitude of the fusion control command is compared with the zero level. When the instantaneous amplitude of the fusion control command is greater than the zero level, the instantaneous amplitude of the fusion control command is retained as the drive preparation signal; when the instantaneous amplitude of the fusion control command is less than or equal to the zero level, the drive preparation signal is set to zero. After the positive amplitude hold processing, the drive preparation signal only contains the signal segments with positive amplitude values ​​from the fusion control command.

[0066] The drive preparation signal is amplified to generate the drive control component. The power amplification process uses linear voltage amplification, multiplying the voltage amplitude of the drive preparation signal by the drive channel power amplification gain to obtain the voltage value of the drive control component. The drive channel power amplification gain is determined based on the ratio of the rated input voltage range of the machine tool functional component execution terminal's drive input port to the maximum amplitude of the fused control command. The value of the drive channel power amplification gain is obtained from the drive module technical manual of the machine tool functional component execution terminal, ensuring that the output voltage range of the drive control component matches the allowable input range of the drive input port.

[0067] In the braking channel, negative amplitude inversion processing is performed on the fusion control command. The negative amplitude inversion processing works as follows: at each sampling moment, the instantaneous amplitude of the fusion control command is compared with the zero level. When the instantaneous amplitude of the fusion control command is less than the zero level, the instantaneous amplitude of the fusion control command is extracted, and the extracted instantaneous amplitude is subjected to absolute value conversion to generate a braking preparation signal; when the instantaneous amplitude of the fusion control command is greater than or equal to the zero level, the braking preparation signal is set to zero. The absolute value conversion process involves multiplying the negative value by -1 to obtain the corresponding positive value. After the negative amplitude inversion processing, the braking preparation signal only contains the positive signal obtained by absolute value conversion of the signal segment with a negative amplitude in the fusion control command.

[0068] The braking preparation signal is amplified to generate the braking control component. The power amplification process uses linear voltage amplification, multiplying the voltage amplitude of the braking preparation signal by the braking channel power amplification gain to obtain the voltage value of the braking control component. The braking channel power amplification gain is determined based on the ratio of the rated input voltage range of the braking input port of the machine tool functional component execution terminal to the maximum amplitude of the fused control command. The value of the braking channel power amplification gain is obtained from the braking module technical manual of the machine tool functional component execution terminal, ensuring that the output voltage range of the braking control component matches the allowable input range of the braking input port.

[0069] In the damping channel, high-frequency component extraction processing is performed on the fused control command. This high-frequency component extraction process uses a high-pass filter to filter out low-frequency components from the fused control command and extracts the high-frequency oscillation components as the damping preparatory signal. The high-pass filter is implemented using a first-order digital high-pass filter, and its difference equation is as follows: in, Indicates the current sampling time sequence number is The damping preparatory signal value output by the high-pass filter; Indicates the previous sampling time number is The damping preparatory signal value output by the high-pass filter; Indicates the current sampling time sequence number is The instantaneous amplitude of the fusion control command for the high-pass filter is input at the moment; Indicates the previous sampling time number is The instantaneous amplitude of the fusion control command for the high-pass filter is input at the moment; This represents the filter coefficients of a first-order digital high-pass filter.

[0070] Filter coefficients The calculation formula is: ,in This indicates the cutoff frequency of the high-pass filter. Indicates the sampling period. Cutoff frequency. The set value is 0.5 times the natural frequency of the machine tool functional component. The natural frequency of the machine tool functional component is obtained through hammer impact modal testing. A pulse excitation is applied to the execution terminal of the machine tool functional component, and the vibration response signal is collected. The first-order natural frequency is extracted from the spectrum of the vibration response signal as the natural frequency of the machine tool functional component. Multiplying the natural frequency of the machine tool functional component by 0.5 yields the cutoff frequency. The setting value is multiplied by 0.5 to ensure that the high-pass filter retains high-frequency oscillation components near and above its natural frequency, while suppressing low-frequency motion components below half of its natural frequency.

[0071] The damping preparation signal is amplitude-scaled to generate the damping adjustment component. The amplitude-scaled process is as follows: the damping preparation signal... The voltage amplitude is multiplied by the damping scaling factor to obtain the voltage value of the damping adjustment component. The damping scaling factor is determined based on the ratio of the maximum allowable input voltage at the damping adjustment input port to the maximum amplitude of the high-pass filter output signal under typical operating conditions. The maximum amplitude under typical operating conditions is taken as the peak recorded value of the damping channel output voltage of the machine tool functional component under no-load high-speed operation.

[0072] The drive control component, braking control component, and damping adjustment component are aligned according to timing. The timing alignment method is as follows: the drive control component, braking control component, and damping adjustment component are written into three first-in-first-out (FIFO) buffers respectively. The writing clock of the three FIFO buffers uses the same sampling clock, and the reading clock of the three FIFO buffers uses the same sampling clock. During reading, the three component data at the same sampling time are retrieved simultaneously from the three FIFO buffers to achieve timing alignment.

[0073] The timing-aligned drive control components are allocated to the drive input port of the machine tool functional component execution terminal, the timing-aligned brake control components are allocated to the brake input port of the machine tool functional component execution terminal, and the timing-aligned damping adjustment components are allocated to the damping adjustment input port of the machine tool functional component execution terminal. After receiving the drive control components, the drive input port of the machine tool functional component execution terminal drives the machine tool functional component to move along the feed direction. After receiving the brake control components, the brake input port applies a reverse braking force. After receiving the damping adjustment components, the damping adjustment input port adjusts the damping force of the adjustable damper. The drive control components, brake control components, and damping adjustment components are output sequentially according to their sampling times, generating the drive signal sequence of the machine tool functional component execution terminal.

[0074] The above description is merely a specific embodiment of this application, but the scope of protection of this application 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 this application should be included within the scope of protection of this application.

Claims

1. An adaptive intelligent control method for machine tool functional components, characterized in that, The method includes: Collect multi-dimensional dynamic response characteristic signals of machine tool functional components during the machining process. The multi-dimensional dynamic response characteristic signals include spindle vibration acceleration sequence, tool holder torque fluctuation sequence, and feed axis displacement deviation sequence. The multidimensional dynamic response feature signal is subjected to joint time-frequency domain feature extraction processing to generate a component operating state feature set consisting of transient impact feature sequence and steady-state cycle feature sequence; A multi-controller collaborative architecture is constructed based on the component operating state feature set. This architecture includes a proportional-integral-derivative control branch and a sliding mode variable structure control branch, comprising: Extract transient impact feature sequences from the component operating state feature set, determine the initial value of the switching gain of the sliding mode variable structure control branch based on the amplitude change rate of the transient impact feature sequences, set the sliding surface parameters based on the initial value of the switching gain, and construct the reaching law function of the sliding mode variable structure control branch; Extract steady-state cyclic feature sequences from the component operating state feature set, determine the proportional coefficient, integral time constant, and derivative time constant of the proportional-integral-derivative control branch based on the periodic stationary characteristics of the steady-state cyclic feature sequences, and construct the transfer function of the proportional-integral-derivative control branch; Establish a parallel connection between the output of the sliding mode variable structure control branch and the output of the proportional-integral-derivative control branch, and set a weight fusion node at the intersection of the parallel connection. Based on the input-output mapping relationship of the weighted fusion nodes, the multi-controller collaborative architecture is constructed. By utilizing the adaptive switching mechanism in the multi-controller collaborative architecture, the proportional-integral-derivative control branch and the sliding mode variable structure control branch are dynamically weighted to generate fused control commands. The fusion control command is decomposed into a drive control component, a braking control component, and a damping adjustment component. A drive signal sequence for the machine tool functional component execution terminal is generated based on the drive control component, the braking control component, and the damping adjustment component.

2. The adaptive intelligent control method for machine tool functional components according to claim 1, characterized in that, The multidimensional dynamic response characteristic signals of the machine tool functional components during the machining process are collected, including: A piezoelectric vibration sensor array is installed on the surface of the machine tool spindle end cover and tool post base, and the vibration acceleration sequence of the spindle is obtained through the piezoelectric vibration sensor array; A Hall current sensor is connected in series on the power supply line of the machine tool spindle drive motor. The Hall current sensor collects the time-domain waveform data of the motor phase current. The time-domain waveform data is processed by three-phase current synthesis to generate the tool holder torque fluctuation sequence. A high-speed counter is connected to the output end of the grating ruler reading head of the machine tool feed axis. The actual position feedback value in each interpolation cycle is recorded by the high-speed counter. The difference between the actual position feedback value and the command position value sent by the CNC system is calculated to generate the feed axis displacement deviation sequence. The spindle vibration acceleration sequence, the tool holder torque fluctuation sequence, and the feed axis displacement deviation sequence are synchronized and aligned according to the sampling timestamps to generate a time-synchronized multidimensional dynamic response feature signal.

3. The adaptive intelligent control method for machine tool functional components according to claim 1, characterized in that, Generate a component operating state feature set consisting of transient impact feature sequences and steady-state cycle feature sequences, including: The main shaft vibration acceleration sequence is subjected to short-time Fourier transform processing to obtain the time spectrum of the vibration signal. The frequency components corresponding to the energy peaks are extracted from the time spectrum as the transient impact feature sequence of the main shaft. Wavelet packet decomposition is performed on the tool holder torque fluctuation sequence. The energy coefficients in each frequency band after decomposition are arranged in order of frequency from low to high to generate a multi-scale energy distribution vector. The frequency band coefficients in the multi-scale energy distribution vector that are higher than the preset energy threshold are extracted as the torque transient impact feature sequence. The autocorrelation function is calculated on the feed axis displacement deviation sequence to extract the main lobe width and side lobe attenuation rate of the autocorrelation function. The steady-state cycle characteristic sequence of the feed system is determined based on the main lobe width and side lobe attenuation rate. The transient impact feature sequence of the main shaft, the transient impact feature sequence of torque, and the steady-state cycle feature sequence are concatenated according to feature dimensions to generate the component operating state feature set.

4. The adaptive intelligent control method for machine tool functional components according to claim 1, characterized in that, The generation of fusion control instructions includes: Obtain the component operating state feature set at the current sampling time, and calculate the state feature deviation degree based on the component operating state feature set at the current sampling time. The state feature deviation degree is the absolute value of the difference between the transient impact feature amplitude at the current time and the transient impact feature amplitude at the previous time. Based on the comparison result between the state characteristic deviation and the preset deviation threshold, the working mode of the adaptive switching mechanism is determined. The working mode includes transient dominant mode, steady-state dominant mode and transitional equilibrium mode. When the adaptive switching mechanism operates in transient-dominant mode, the participation weight of the sliding mode variable structure control branch is set to a first weight value, and the participation weight of the proportional-integral-derivative control branch is set to a second weight value, wherein the first weight value is greater than the second weight value. When the adaptive switching mechanism operates in steady-state dominant mode, the participation weight of the proportional-integral-derivative control branch is set to the third weight value, and the participation weight of the sliding mode variable structure control branch is set to the fourth weight value, wherein the third weight value is greater than the fourth weight value. When the adaptive switching mechanism operates in transition equilibrium mode, the participation weights of the proportional integral derivative control branch and the sliding mode variable structure control branch are linearly adjusted according to the difference ratio between the state characteristic deviation degree and the preset deviation degree threshold, so that the sum of the participation weights of the two branches remains constant. Based on the output values ​​of the proportional-integral-derivative control branch and the sliding mode variable structure control branch, as well as the participation weights of the two branches, a weighted summation calculation is performed to generate the fusion control command.

5. The adaptive intelligent control method for machine tool functional components according to claim 1, characterized in that, The fused control command is decomposed into a drive control component, a braking control component, and a damping adjustment component. A drive signal sequence for the machine tool functional component execution terminal is generated based on the drive control component, the braking component, and the damping adjustment component, including: The fusion control command is input to a three-channel resolver, which includes a drive channel, a braking channel, and a damping channel. In the drive channel, the fusion control command is subjected to positive amplitude holding processing, and the signal segment with a positive amplitude in the fusion control command is extracted as the drive preparation signal. The drive preparation signal is then subjected to power amplification processing to generate the drive control component. In the braking channel, the fusion control command is subjected to negative amplitude inversion processing, the signal segment with negative amplitude in the fusion control command is extracted and absolute value conversion processing is performed to generate a braking preparation signal, and the braking preparation signal is subjected to power amplification processing to generate the braking control component. In the damping channel, high-frequency component extraction processing is performed on the fusion control command. Low-frequency components in the fusion control command are filtered out by a high-pass filter, and high-frequency oscillation components in the fusion control command are extracted as damping preparation signals. The damping preparation signals are then subjected to amplitude scaling processing to generate the damping adjustment components. The drive control component, the brake control component, and the damping adjustment component are aligned according to their timing sequence and then distributed to the drive input port, brake input port, and damping adjustment input port of the machine tool functional component execution terminal, respectively, to generate the drive signal sequence.

6. The adaptive intelligent control method for machine tool functional components according to claim 2, characterized in that, When synchronizing and aligning the spindle vibration acceleration sequence, the tool holder torque fluctuation sequence, and the feed axis displacement deviation sequence according to the sampling timestamp, the sampling clock of the tool holder torque fluctuation sequence is used as the reference clock. Linear interpolation resampling is then performed on the spindle vibration acceleration sequence and the feed axis displacement deviation sequence respectively, so that the three sets of sequences have the same sampling frequency and time coordinate origin.

7. The adaptive intelligent control method for machine tool functional components according to claim 3, characterized in that, The wavelet packet decomposition process has 5 decomposition layers and uses the Daubechies wavelet basis function. The preset energy threshold is set to 1.5 times the average energy of each frequency band.

8. The adaptive intelligent control method for machine tool functional components according to claim 1, characterized in that, The establishment of a parallel connection between the output of the sliding mode variable structure control branch and the output of the proportional-integral-derivative (PI-DE) control branch, and the setting of a weight fusion node at the intersection of the parallel connection, includes: The output of the sliding mode variable structure control branch is connected to the first input of the weighted fusion node through a first signal conditioning circuit. The first signal conditioning circuit includes a first limiter and a first impedance matching network. The output of the proportional-integral-derivative control branch is connected to the second input of the weighted fusion node through a second signal conditioning circuit, which includes a second limiter and a second impedance matching network. Inside the weight fusion node, a first multiplier, a second multiplier, and an adder are set. The sliding mode variable structure control branch output by the adaptive switching mechanism is included in the weight input to the first multiplier input terminal of the first multiplier. The output value of the sliding mode variable structure control branch after processing by the first signal conditioning circuit is input to the second multiplier input terminal of the first multiplier. The proportional-integral-derivative control branch output by the adaptive switching mechanism is included in the weight input to the first multiplier input terminal of the second multiplier. The output value of the proportional-integral-derivative control branch after processing by the second signal conditioning circuit is input to the second multiplier input terminal of the second multiplier. The product output of the first multiplier and the product output of the second multiplier are respectively connected to the two adder inputs of the adder, and the fusion control command is obtained at the output of the adder.

9. The adaptive intelligent control method for machine tool functional components according to claim 4, characterized in that, The state characteristic deviation is the ratio of the absolute value of the difference between the transient impact characteristic amplitude at the current moment and the transient impact characteristic amplitude at the previous moment to the transient impact characteristic amplitude at the previous moment, and the preset deviation threshold is set to 0.15.

Images