Numerical control machine tool vibration suppression self-adaptive control system and method

Abstract

The invention discloses a numerical control machine tool vibration suppression self-adaptive control system and method, and relates to the technical field of intelligent vibration control, and the method comprises the steps: building a current-modal dynamic mapping relation between an electrical drive characteristic and a mechanical structure modal response based on a vibration state feature vector; real-time machine tool modal state vectors are obtained through online recursive estimation; the machine tool modal state vector is used as a boundary condition to be injected into the virtual prediction model; solving to obtain an optimal control parameter sequence and a corresponding vibration prediction result in a future time domain by taking multi-performance index collaborative optimization as a target through a rolling optimization control strategy; translating the control parameter sequence into a specific execution instruction and issuing the specific execution instruction; monitoring an execution process in real time and recording actual response data; according to the method, the accuracy of vibration prediction and the real-time performance of control response are improved by establishing the dynamic mapping relation from the electromagnetic driving characteristics of the servo main shaft and the feed motor to the dynamic response of the machine tool structure.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent vibration control, in particular to a vibration suppression adaptive control system and method for a numerical control machine tool. BACKGROUND

[0002] With the development of high-end manufacturing equipment towards high speed, high precision and high dynamic performance, the vibration problem of numerical control machine tools gradually becomes an important factor affecting the machining quality and equipment life. The existing vibration suppression control technology is mainly based on parameter adjustment or additional vibration compensation control of the machine tool servo system. By collecting the current, speed and acceleration signals of the servo spindle and the feed motor, the vibration characteristics of the machine tool structure are identified using frequency spectrum analysis or time domain modeling method, and a feedforward compensation or filtering algorithm is introduced in the control loop to realize the suppression of vibration at a specific frequency.

[0003] The limitations of the prior art mainly lie in that the traditional numerical control machine tool vibration control method cannot realize real-time coupling modeling and adaptive control of the dynamic response of electromagnetic drive and mechanical structure. For example, when the machine tool runs at high speed cutting or multi-axis linkage, the coupling relationship between the electrical drive characteristics and the mechanical structure modal will dynamically drift with the change of working conditions, while the traditional method relies on static model or fixed control parameters, which is difficult to reflect this change in time, resulting in lag of vibration prediction, inaccurate control instruction response, and unstable vibration suppression effect. SUMMARY

[0004] In view of the above existing problems, the present application is proposed.

[0005] Therefore, the present application provides a numerical control machine tool vibration suppression adaptive control method to solve the problem of vibration prediction accuracy and control effect decline caused by the inability to model and correct the coupling characteristics of electrical drive and mechanical structure in dynamic operation.

[0006] To solve the above technical problems, the present application provides the following technical solutions:

[0007] In a first aspect, the present application provides a numerical control machine tool vibration suppression adaptive control method, which comprises,

[0008] extracting frequency domain and time domain features from the multi-source sensing signals of the spindle-feed motor to generate a vibration state feature vector;

[0009] establishing a current-modal dynamic mapping relationship between the electrical drive characteristics and the mechanical structure modal response based on the vibration state feature vector; and obtaining a real-time machine tool modal state vector through online recursive estimation;

[0010] The machine tool modal state vector is injected as a boundary condition into a virtual prediction model; and a rolling optimization control strategy is used to solve for an optimal control parameter sequence and corresponding vibration prediction results in a future time domain with the goal of multi-performance index collaborative optimization;

[0011] The control parameter sequence is translated into specific execution instructions and distributed; the execution process is monitored in real time and actual response data is recorded;

[0012] The vibration response residual is calculated based on the actual response data and the vibration prediction results; and the current-modal dynamic mapping relationship and the virtual prediction model are corrected online according to the vibration response residual.

[0013] As a preferred scheme of the numerical control machine tool vibration suppression adaptive control method, wherein:

[0014] The method for generating the vibration state feature vector comprises,

[0015] The mean value subtraction processing for eliminating direct current offset and the low-pass filtering processing for suppressing frequency spectrum aliasing are performed on the multi-source sensing signals of the spindle-feeding shaft motor;

[0016] A fixed-length sliding window is used to perform continuous segmentation on the preprocessed multi-source sensing signals; each segmented multi-source sensing signal is converted into a corresponding frequency domain amplitude spectrum to extract frequency domain features, and the position feedback signal of the feeding motor is compared with the command position signal output by the servo controller to obtain a trajectory deviation to extract time domain features;

[0017] The frequency domain features and the time domain features are spliced and combined and numerically normalized to generate the vibration state feature vector.

[0018] As a preferred scheme of the numerical control machine tool vibration suppression adaptive control method, wherein: the method for establishing the current-modal dynamic mapping relationship comprises,

[0019] The frequency domain features are separated from the vibration state feature vector to extract frequency domain response features reflecting electromagnetic drive characteristics, forming an electrical drive characteristic; and the time domain features are separated from the vibration state feature vector to extract time domain response features reflecting mechanical dynamics characteristics, forming a mechanical structure modal response;

[0020] The electrical drive characteristic and the mechanical structure modal response are time-aligned and synchronously paired to form an electrical-mechanical corresponding time series data set;

[0021] Based on the change rule of the electrical-mechanical corresponding time series data set in the time series, a multivariate correlation relationship is established with the electrical drive characteristic as the input and the mechanical structure modal response as the output; and the correlation weight of the electrical drive characteristic and the mechanical structure modal response under different time delays is calculated to construct a dynamic influence matrix;

[0022] The adjustable parameters in the dynamic influence matrix for representing the strength and direction of the influence of the electrical drive characteristics on the modal response of the mechanical structure are defined as correlation parameters; the correlation parameters are adaptively adjusted in a recursive updating manner, so that the multi-variable correlation is continuously convergent in the time sequence, to establish a current-mode dynamic mapping relationship.

[0023] As a preferred scheme of the vibration suppression adaptive control method for the numerical control machine tool according to the application, the method for obtaining the real-time machine tool modal state vector through online recursive estimation comprises,

[0024] Based on the current-mode dynamic mapping relationship, the electrical drive characteristics and the modal response of the mechanical structure at the current time are received, are synchronously paired according to the time stamp, and an electrical-mechanical time sequence input set is constructed;

[0025] Based on the electrical-mechanical time sequence input set and the current-mode dynamic mapping relationship, the machine tool modal parameter estimation value at the current time is calculated, and an initial machine tool modal state vector is generated;

[0026] The initial machine tool modal state vector is compared with the machine tool modal state vector at the previous time, the correlation parameters are dynamically adjusted according to the difference, the mapping accuracy is corrected, and a corrected current-mode dynamic mapping relationship is obtained;

[0027] Based on the corrected current-mode dynamic mapping relationship and the electrical-mechanical time sequence input set, online recursive estimation calculation is performed, and the machine tool modal state vector is obtained in real time.

[0028] As a preferred scheme of the vibration suppression adaptive control method for the numerical control machine tool according to the application, the method for solving the optimal control parameter sequence in the future time domain and the corresponding vibration prediction result comprises,

[0029] The machine tool modal state vector is input into a virtual prediction model; the dynamic boundary condition of the machine tool is determined according to the structural modal characteristic parameters in the machine tool modal state vector; and the dynamic boundary condition is assigned to the initial state variable of the virtual prediction model;

[0030] Based on the multi-step forward simulation architecture of the virtual prediction model, the current control instruction and the historical operation data are fused in real time, the state evolution trajectory in the future N control periods is deduced under the kinematic and dynamic constraint conditions through the time domain iterative solution of the machine tool dynamics equation, the coupled vibration modal response of the spindle-feed shaft motor is synchronously calculated, and a machine tool future state prediction sequence is formed;

[0031] Based on the machine tool future state prediction sequence, a collaborative optimization objective function describing the comprehensive performance index is constructed; and the machine tool modal state vector is taken as a constraint condition;

[0032] Based on the collaborative optimization objective function, a rolling optimization control strategy is adopted for real-time solution, in each optimization period, a plurality of time domain steps are slidingly predicted in the future, the electrical drive control variables of the spindle-feed axis motor are dynamically adjusted, the collaborative optimization objective function is gradually converged, and the optimal control parameter sequence is obtained;

[0033] The optimal control parameter sequence is re-input into the virtual prediction model for dynamic simulation, and the vibration prediction result in the corresponding time domain is output.

[0034] As a preferred scheme of the vibration suppression adaptive control method of the numerical control machine tool, wherein: the method of translating the control parameter sequence into specific execution instructions and issuing includes,

[0035] Based on the optimal control parameter sequence, the corresponding electrical drive control variables of the spindle-feed axis motor are extracted for each time domain step; and according to the machine tool control interface protocol and the servo drive logic, the mapping is translated into a set of machine tool executable control instructions to generate corresponding execution instruction data packets;

[0036] The execution instruction data packet is issued to the machine tool controller, and the spindle-feed axis motor is driven by the controller to execute the corresponding control.

[0037] As a preferred scheme of the vibration suppression adaptive control method of the numerical control machine tool, wherein: the method of real-time monitoring the execution process and recording the actual response data includes,

[0038] In the execution instruction data packet process, the machine tool running state signal of the spindle-feed axis motor is monitored in real time to form an execution monitoring original data set;

[0039] Based on the execution monitoring original data set, the electrical drive characteristics and the mechanical structure modal response are extracted, the machine tool modal state vector in the control execution process of the machine tool is calculated, and the actual response data under the corresponding time domain step is generated.

[0040] As a preferred scheme of the vibration suppression adaptive control method of the numerical control machine tool, wherein: the method of calculating the vibration response residual based on the actual response data and the vibration prediction result includes,

[0041] The actual response data and the vibration prediction result are matched according to the same time domain step index, and the time stamp is aligned to form a prediction-measurement comparison data set;

[0042] The machine tool modal state vector under the corresponding time domain step in the prediction-measurement comparison data set is element-by-element differentiated to obtain a modal response difference sequence;

[0043] Based on the modal response difference sequence, a comprehensive difference index is calculated by using a weighted mean square error criterion, and the comprehensive difference index is defined as the vibration response residual corresponding to the time domain step.

[0044] As a preferred scheme of the vibration suppression adaptive control method of the numerical control machine tool, the method of online parameter correction of the current-modal dynamic mapping relationship and the virtual prediction model according to the vibration response residual comprises,

[0045] Based on the vibration response residual, the modal response difference sequence at each time domain step is analyzed, and a residual feature vector reflecting the deviation of the electrical driving characteristics and the modal response deviation of the mechanical structure is extracted; the corresponding weight is set according to the contribution of each modal parameter in the machine tool modal state vector to the overall vibration energy, and the residual feature vector is weighted to form a weighted residual feature set;

[0046] The weighted residual feature set is input into the current-modal dynamic mapping relationship, the recursive adjustment of the associated parameters is performed according to the residual feature vector, the dynamic influence matrix between the electrical driving characteristics and the modal response of the mechanical structure is re-converged in the time sequence, and the corrected current-modal dynamic mapping relationship is obtained;

[0047] Based on the corrected current-modal dynamic mapping relationship and the vibration response residual, the machine tool modal state vector in the virtual prediction model is re-estimated; and the parameters of the machine tool dynamics equation in the virtual prediction model are updated with the new machine tool modal state vector.

[0048] In the second aspect, the present application provides a numerical control machine tool vibration suppression adaptive control system, comprising,

[0049] The feature extraction module is used for frequency domain and time domain feature extraction of the multi-source sensing signals of the spindle-feeding shaft motor, and generates a vibration state feature vector.

[0050] The mapping modeling module is used for establishing a current-modal dynamic mapping relationship between the electrical driving characteristics and the modal response of the mechanical structure based on the vibration state feature vector, and obtaining a real-time machine tool modal state vector through online recursive estimation.

[0051] The prediction optimization module is used for injecting the machine tool modal state vector into the virtual prediction model as a boundary condition, and solving an optimal control parameter sequence and a corresponding vibration prediction result in the future time domain through a multi-performance index collaborative optimization target by using a rolling optimization control strategy.

[0052] The execution monitoring module is used for translating the control parameter sequence into specific execution instructions and issuing; the execution process is monitored in real time and the actual response data is recorded.

[0053] An online correction module is configured to calculate vibration response residuals based on actual response data and vibration prediction results, and to perform online parameter correction on the current-mode dynamic mapping relationship and the virtual prediction model according to the vibration response residuals.

[0054] The present application has the following advantages: by establishing a dynamic mapping relationship between the electromagnetic drive characteristics of the servo main shaft and the feed motor and the dynamic response of the machine tool structure, real-time coupling modeling of the electrical drive characteristics and the mechanical structure modal is achieved, and an adaptive vibration control model is constructed based on the mapping relationship; by introducing dynamic response feature recognition and optimal control parameter self-learning mechanism, the control instruction can be updated in real time during the operation of the machine tool, so that the control strategy is dynamically adjusted according to the running state, thereby overcoming the problems of control lag and static modeling in traditional methods. The technology not only improves the accuracy of vibration prediction and the real-time performance of control response, but also improves the system stability and machining precision under complex working conditions such as high-speed cutting and multi-axis linkage, and realizes active suppression and performance adaptive optimization of machine tool vibration. BRIEF DESCRIPTION OF DRAWINGS

[0055] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0056] Figure 1 The flowchart of the adaptive control method for vibration suppression of the numerical control machine tool in the present application;

[0057] Figure 2 The schematic diagram of the adaptive control system for vibration suppression of the numerical control machine tool in the present application;

[0058] Figure 3 The flowchart of constructing the current-mode dynamic mapping relationship in the present application;

[0059] Figure 4 The flowchart of outputting the vibration prediction result in the present application. DETAILED DESCRIPTION

[0060] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application will be described in detail in conjunction with the drawings of the specification.

[0061] In the following description, many specific details are set forth in order to provide a thorough understanding of the present application, but the present application can also be implemented in other ways different from those described herein, and those skilled in the art can make similar generalizations without departing from the connotation of the present application, therefore the present application is not limited to the specific embodiments disclosed below.

[0062] Second, the "one embodiment" or "an embodiment" referred to herein means a specific feature, structure, characteristic, or combination of features and / or characteristics described herein that can be included in at least one implementation of the present application. The various appearances of "in one embodiment" or "an embodiment" in the specification are not necessarily all referring to the same embodiment.

[0063] Referring to Figure 1 , Figure 2 , Figure 3 and Figure 4 , one embodiment of the present application provides a vibration suppression adaptive control method for a numerical control machine tool, comprising the following steps:

[0064] The method for generating the vibration state feature vector comprises:

[0065] The multi-source sensing signals of the spindle-feeding shaft motor are subjected to mean value subtraction processing for eliminating DC offset and low-pass filtering processing for suppressing spectral aliasing.

[0066] It should be noted that the multi-source sensing signals are collected by a current sensor for the driving current signals of the spindle-feeding shaft motor, by a voltage sensor for the power supply voltage signals, and by an encoder for the motor rotor position signals;

[0067] Further, the multi-source sensing signals of the spindle-feeding shaft motor are subjected to mean value subtraction processing for eliminating DC offset, the arithmetic mean value of the multi-source sensing signals within a complete signal period or a fixed time window is calculated, and the mean value is subtracted from each instantaneous sampling value of the multi-source sensing signals; the multi-source sensing signals of the spindle-feeding shaft motor are subjected to low-pass filtering processing for suppressing spectral aliasing, a low-pass filter with a cutoff frequency lower than one-half of the sampling frequency of the multi-source sensing signals is used to filter the multi-source sensing signals after the mean value subtraction processing, and the frequency components higher than the cutoff frequency in the multi-source sensing signals are filtered out.

[0068] The preprocessed multi-source sensing signals are subjected to continuous segmentation by using a fixed-length sliding window; each segmented multi-source sensing signal is converted into a corresponding frequency domain amplitude spectrum to extract frequency domain features, and the position feedback signal of the feeding motor is compared with the command position signal output by the servo controller to obtain a trajectory deviation to extract time domain features.

[0069] It should be noted that the fixed-length sliding window is exemplarily a 200-millisecond window, which is placed from the starting point of the preprocessed multi-source sensing signals, and all the multi-source sensing signal sampling points contained in the sliding window are divided into one data segment; the window is slid by a fixed step length, for example, 50 milliseconds, to obtain the next data segment, and the operation is repeated until the entire signal is covered, and the continuous segmentation is completed.

[0070] Apply Fast Fourier Transform to each segmented multi-source sensor signal data segment to transform the multi-source sensor signal from time domain representation to frequency domain representation. Calculate the amplitude value of each frequency component in the frequency domain representation to form the frequency domain amplitude spectrum corresponding to the data segment. Extract the maximum amplitude value or average amplitude value within a specific frequency band from the frequency domain amplitude spectrum as frequency domain features.

[0071] Within the same time interval, the actual value of the position feedback signal of the feed motor is read, the set value of the command position signal output by the servo controller is read, the absolute value of the difference between the position feedback signal of the feed motor and the command position signal output by the servo controller at each sampling moment is calculated, the root mean square value of all differences within the time interval is calculated as the trajectory deviation, and the trajectory deviation value is used as the time domain feature.

[0072] The frequency domain features and time domain features are concatenated and combined, and the values ​​are normalized to generate a vibration state feature vector.

[0073] It should be noted that all the numerical elements of the frequency domain features and all the numerical elements of the time domain features are concatenated into a new one-dimensional array in the order of frequency domain features first and time domain features last, thus completing the splicing and combination.

[0074] The concatenated one-dimensional array is subjected to numerical normalization using a min-max normalization method: for each numerical element in the array, its feature category is determined. If the element belongs to the frequency domain feature, the frequency domain feature normalization parameter is used, with the exemplary minimum value set to zero and the maximum value set to ten. If the element belongs to the time domain feature, the time domain feature normalization parameter is used, with the exemplary minimum value set to zero and the maximum value set to one hundred micrometers. The normalization result for each numerical element is calculated. The numerical elements in all the one-dimensional arrays after normalization are arranged in their original order to form a vibration state feature vector.

[0075] Methods for establishing the current-mode dynamic mapping relationship include:

[0076] Existing technologies mainly rely on independent analysis of electrical or mechanical signals, failing to effectively establish the dynamic coupling relationship between the two. When machine tool operating conditions change, the correlation characteristics between electrical drive and mechanical modes drift, resulting in poor adaptability and reduced control effect of vibration suppression methods based on fixed parameters, making it impossible to achieve precise vibration suppression.

[0077] Frequency domain features are separated from the vibration state feature vector, and frequency domain response features reflecting electromagnetic drive characteristics are extracted to form electrical drive characteristics; time domain features are separated from the vibration state feature vector, and time domain response features reflecting mechanical dynamic characteristics are extracted to form mechanical structure modal response.

[0078] Specifically, all numerical elements of the vibration state feature vector are read in the order they were generated, and the first few frequency domain feature numerical elements are extracted to form a frequency domain feature set. The amplitude of the frequency components corresponding to each numerical element in the frequency domain feature set is compared, and the amplitude proportion sequence of each frequency component is calculated. The amplitude average of the dominant frequency component in the interval with the largest amplitude proportion and its adjacent frequency bands is used as the electromagnetic drive frequency band feature. The amplitude concentration parameter is calculated based on the concentration of amplitude changes in the frequency domain feature set. The amplitude concentration parameter is defined as the ratio of the dominant frequency amplitude to the average amplitude of all frequencies. The amplitude concentration parameter is used to characterize the electrical drive characteristics.

[0079] The temporal feature elements arranged later are extracted individually to form a temporal feature set. The change sequence of the root mean square value of trajectory deviation in the temporal feature set is calculated to determine the rate of change of the root mean square value of trajectory deviation. Based on the fluctuation amplitude when the rate of change of the root mean square value of trajectory deviation exceeds a preset modal drift threshold, the modal response features in the corresponding time interval are extracted. The modal drift threshold is set for example to be twice the static average level of the root mean square value of trajectory deviation. When the rate of change of the root mean square value of trajectory deviation is greater than this modal drift threshold, the vibration amplitude and phase shift of that period are recorded as the modal response of the mechanical structure.

[0080] The electrical drive characteristics and mechanical structure modal responses are time-aligned and synchronized to form an electrical-mechanical corresponding time series dataset.

[0081] Specifically, timestamp information is read from the electrical drive characteristics and the mechanical structure modal response respectively, and a time series index table is established. The higher of the sampling frequencies of the electrical drive characteristics and the mechanical structure modal response is used as a unified sampling benchmark. The time axis is linearly interpolated with a unified step size to generate the resampled electrical drive characteristic sequence and mechanical structure modal response sequence.

[0082] The generated resampled electrical drive characteristic sequence and mechanical structure modal response sequence are divided into fixed windows, with an example window length of 200 milliseconds and a step size of 50 milliseconds. Within each time window, values ​​are paired according to the timestamp matching principle to form a one-to-one correspondence between electrical drive characteristics and mechanical structure modal responses. The pairing results are validated. When the time interval error is less than the example of 1 millisecond, local interpolation compensation is performed; when it exceeds 1 millisecond, the data for that time period is discarded. Finally, all time window pairing samples that pass the validation are collected into an electrical-mechanical corresponding time series dataset.

[0083] Based on the variation patterns of the electrical-mechanical corresponding time series dataset in the time series, a multivariate correlation relationship is established with electrical drive characteristics as input and mechanical structure modal response as output; and the correlation weights between electrical drive characteristics and mechanical structure modal response under different time delays are calculated to construct a dynamic influence matrix.

[0084] Specifically, the electrical-mechanical time-series datasets are arranged chronologically. The electrical drive characteristic values ​​at each sampling moment constitute the independent variable sequence, while the mechanical structure modal response values ​​at the corresponding moment and several subsequent sampling moments constitute the dependent variable sequence. A multiple linear regression method is used to fit the relationship between the independent and dependent variable sequences, and the regression coefficient matrix is ​​solved. This regression coefficient matrix represents the mathematical relationship between the immediate and delayed effects of electrical drive characteristics on the mechanical structure modal response, thus establishing a multivariate correlation.

[0085] Furthermore, a set of time delay parameters is set, starting from zero and increasing in steps with the sampling period. Exemplary time delay settings include 0 milliseconds, 1 millisecond, 2 milliseconds, up to 10 milliseconds. For each specific time delay value, the electrical drive characteristic sequence is shifted backward along the time axis by that time delay value, resulting in a time-shifted electrical drive characteristic sequence. The Pearson correlation coefficient between the time-shifted electrical drive characteristic sequence and the original mechanical structure modal response sequence is calculated; the absolute value of this coefficient is the correlation weight for that specific time delay. By iterating through all the set time delay parameters, a set of time delay-weight correspondences is obtained. Using time delay as the row index and different components of the electrical drive characteristics (such as amplitude concentration parameters, dominant frequency amplitude, etc.) as column indices, the calculated correlation weights for each time delay are filled into the corresponding positions in the matrix, thus constructing the dynamic influence matrix.

[0086] It should be noted that the expression for the Pearson correlation coefficient between the time-shifted electrical drive characteristic sequence and the original mechanical structure modal response sequence is as follows:

[0087]

[0088] Where r is the Pearson correlation coefficient, n is the total number of sampling points within the window, and X i It is the value of the time-shifted electrical drive characteristic sequence at the i-th sampling point. Y is the average value of the time-shifted electrical drive characteristic sequence within the calculation window. i It is the value of the original mechanical structure modal response sequence at the i-th sampling point. It is the average value of the original mechanical structure modal response sequence within the calculation window, and i is the index variable of the sampling points within the window.

[0089] The adjustable parameters in the dynamic influence matrix used to characterize the intensity and direction of the influence of electrical drive characteristics on the modal response of mechanical structures are defined as correlation parameters. The correlation parameters are adaptively adjusted by a recursive update method so that the multivariate correlation relationship continues to converge in the time series, thereby establishing a current-modal dynamic mapping relationship.

[0090] It should be noted that the recursive update method is based on the dynamic changes of the continuous sampling window in the time series, and uses the sliding window recursive least squares (RLS) algorithm or the adaptive gradient descent algorithm to gradually correct the associated parameters;

[0091] Specifically, at the arrival of each new sampling window, the electrical drive characteristics of the current window are used as the input vector, and the mechanical structure modal response is used as the target vector. The prediction error is calculated, and the correlation parameters from the previous step are adjusted according to the error direction and magnitude. For example, when using the recursive least squares algorithm, the update of the correlation parameters can be expressed as:

[0092]

[0093] Where, θ t It is the correlation parameter at time t, θ t-1 J is the correlation parameter from the previous time step. t It is the gain vector at time t, y t The observed modal response of the mechanical structure at time t. It is the transpose of the electrical drive characteristic input vector at time t, where T is the matrix transpose operation and t is the time index variable;

[0094] Through a recursive update process, the mapping relationship between electrical drive characteristics and mechanical modal response can be automatically corrected when operating conditions change or structural characteristics drift, dynamically reflecting the time-varying influence of electrical drive on mechanical modal response.

[0095] After multiple iterations, when the prediction error converges and the change amplitude of the associated parameters is lower than the change threshold, the convergence threshold of the exemplary prediction error can be taken as 0.0001 to 0.00001, and the change amplitude threshold can be taken as 0.001 to 0.0001. The resulting set of associated parameters constitutes a stable current-modal dynamic mapping relationship. This current-modal dynamic mapping relationship can reflect the dynamic coupling effect of the machine tool drive current waveform change on the vibration response of the mechanical structure in real time, providing a quantitative basis for the adaptive parameter adjustment of the subsequent vibration suppression controller.

[0096] By constructing an electrical-mechanical time-series dataset and calculating the correlation weights under different time delays to form a dynamic influence matrix, the dynamic effect of electrical drive characteristics on the modal response of mechanical structures is accurately described. A recursive update method is used to continuously adjust the correlation parameters in the dynamic influence matrix, ensuring that the current-modal dynamic mapping relationship closely approximates the actual dynamic characteristics of the machine tool in real time. This provides an accurate basis for adaptive control and effectively improves the accuracy and stability of vibration suppression.

[0097] Methods for obtaining real-time machine tool modal state vectors through online recursive estimation include:

[0098] Based on the current-mode dynamic mapping relationship, the electrical drive characteristics and mechanical structure modal response at the current moment are received, and synchronous pairing is performed according to the timestamp to construct an electrical-mechanical timing input set.

[0099] It should be noted that the electrical-mechanical timing input set consists of electrical drive characteristic vectors and mechanical structure modal response vectors collected within the same time window. The sampling timestamps undergo uniform interpolation and error correction to ensure time alignment accuracy. For example, linear interpolation compensation is performed when the time pairing error is less than 1 millisecond; data for periods exceeding 1 millisecond are discarded, thus ensuring the synchronization and timing continuity of the input set.

[0100] Based on the electrical-mechanical timing input set and the current-modal dynamic mapping relationship, the estimated values ​​of the machine tool modal parameters at the current moment are calculated, and the initial machine tool modal state vector is generated.

[0101] It should be noted that the expression for calculating the estimated values ​​of the machine tool modal parameters at the current moment is:

[0102]

[0103] in, x is the estimated value of the machine tool modal parameters at time t. t It is the electrical-mechanical timing input vector at time t;

[0104] After obtaining the estimated values ​​of the machine tool modal parameters at the current moment, the modal parameters of each order are normalized according to their physical meaning, and vector-level splicing is performed according to the preset modal feature arrangement order (such as from low order to high order) to form a comprehensive feature vector containing the frequency, damping ratio and response amplitude of each order, which is the initial machine tool modal state vector.

[0105] The initial machine tool modal state vector is compared with the machine tool modal state vector at the previous moment. The correlation parameters are dynamically adjusted based on the differences to correct the mapping accuracy and obtain the corrected current-modal dynamic mapping relationship.

[0106] Specifically, the initial machine tool modal state vector is compared with the machine tool modal state vector of the previous moment. When the difference exceeds the dynamic correction threshold (0.001 for example), the gain vector or learning rate in the recursive update algorithm is fine-tuned according to the direction of the difference to enhance the adaptive ability to short-term dynamic changes. Thus, the self-correction of the correlation parameters is completed under the same recursive update framework to obtain the adjusted correlation parameters. These adjusted parameters are then used to replace the original correlation parameters and re-substituted into the current-modal dynamic mapping relationship to obtain the corrected current-modal dynamic mapping relationship.

[0107] Based on the corrected current-modal dynamic mapping relationship and the electrical-mechanical timing input set, online recursive estimation calculation is performed to obtain the machine tool modal state vector in real time.

[0108] Specifically, based on the corrected current-mode dynamic mapping relationship, the continuously arriving electrical-mechanical timing input set is received. When a new sampling window arrives, the electrical drive characteristic vector of the current window is used as input, the modified correlation parameter set is used to predict the modal state, and the prediction error is calculated based on the real-time collected mechanical structure modal response values.

[0109] In the recursive estimation process, the covariance matrix of the input vector is first calculated based on the electrical-mechanical time-series input set to characterize the temporal correlation between electrical drive characteristics and mechanical modal response; at the same time, a forgetting factor (exemplarily 0.98 to 0.995) is set to control the influence weight of historical samples on the current estimation result, so that the estimation result can take into account both short-term dynamic changes and long-term stability.

[0110] When electrical drive characteristics and mechanical structure modal response data arrive within the current sampling window, the values ​​of the associated parameters are dynamically adjusted according to the direction and magnitude of the current prediction error to correct the accuracy of the current-modal dynamic mapping relationship. The gain vector is updated based on the changes in the covariance matrix and forgetting factor of the input vector, ensuring that the parameters remain stable and convergent at different processing stages.

[0111] Through the above recursive update process, the corrected current-mode dynamic mapping relationship can be continuously and adaptively adjusted over time, thereby generating the updated machine tool modal state vector in real time.

[0112] Methods for obtaining the optimal control parameter sequence and corresponding vibration prediction results in the future time domain include,

[0113] Existing vibration control methods for CNC machine tools mostly rely on static models or single-step feedback strategies, which cannot accurately reflect the nonlinear changes and multi-axis coupling effects of the machine tool structure during dynamic operation. This results in lag in vibration prediction and untimely updates to control parameters. Furthermore, traditional methods lack a rolling prediction mechanism for the future time domain, making it difficult to achieve forward-looking control of machine tool vibration trends under high-speed cutting conditions.

[0114] Input the machine tool modal state vector into the virtual prediction model; determine the dynamic boundary conditions of the machine tool based on the structural modal characteristic parameters in the machine tool modal state vector; and assign the dynamic boundary conditions to the initial state variables of the virtual prediction model.

[0115] It should be noted that the virtual prediction model can employ a multibody dynamics model or a finite element dynamics model that includes the spindle, feed axis, and machine tool support structure. It describes the dynamic response characteristics of the machine tool under multi-axis linkage and cutting loads through state equations. Based on the structural modal characteristic parameters (including natural frequencies, damping ratios, mode shapes, and response amplitudes) of each order in the machine tool modal state vector, the corresponding modal stiffness matrix and damping matrix are calculated to determine the dynamic boundary conditions of the machine tool. These determined dynamic boundary conditions are then used as constraint inputs to the initial state variables of the virtual prediction model, ensuring that the virtual prediction model reflects the consistency between the current machine tool structural dynamic characteristics and the actual working state.

[0116] It should be noted that the stiffness matrix and damping matrix for calculating the e-th mode are expressed as follows:

[0117] K m (e)=(2πf e ) 2 M e I m (e)=2ζ e (2πf e M e ;

[0118] Among them, K m (e) is the equivalent stiffness coefficient of the e-th mode, f e M is the natural frequency of the e-th order mode. e It is the e-th order equivalent modal mass, I m (e) is the equivalent damping coefficient of the e-th mode, ζ e is the e-th order damping ratio, m is the total modal order of the machine tool structure, and e is the index variable of the modal order of the machine tool structure;

[0119] It should be noted that K m (e) and I m (e) are the scalar parameters of a single-order mode. By combining the stiffness and damping of each order mode in a diagonalized manner, the overall modal stiffness matrix and modal damping matrix are formed.

[0120] Based on a multi-step forward simulation architecture using a virtual prediction model, the system integrates current control commands and historical operating data in real time. By solving the machine tool dynamics equations in the time domain, it extrapolates the state evolution trajectory over the next N control cycles under kinematic and dynamic constraints. Simultaneously, it calculates the coupled vibration modal response of the spindle-feed axis motors, forming a predictive sequence of the machine tool's future state.

[0121] Specifically, within each control cycle, the virtual prediction model simultaneously receives the current electrical drive control command and recorded historical operating data as input boundary conditions, substitutes their load terms into the machine tool dynamics equations, and performs discretized time-domain iterative solution. The machine tool dynamics equations can be represented as follows:

[0122]

[0123] Among them, M s It is the equivalent mass matrix of the machine tool in the modal coordinate system. It is the modal acceleration vector at time t, C s It is the modal damping matrix. It is the velocity vector of the mode at time t. It is the corrected modal stiffness matrix, q s F(t) is the modal displacement vector at time t, F s (t) is the equivalent force vector formed by the electrical drive input and the external cutting load at time t;

[0124] During the solution process, the continuous-time dynamic equations are discretized into time-step differential forms. The displacement, velocity, and acceleration responses at each time step are calculated step by step using numerical integration algorithms (such as the Newmark-β method or the fourth-order Runge-Kutta method). Based on the state variables (modal displacement, velocity, and acceleration) and input load of the previous time step, the dynamic response at the current moment is iteratively solved to achieve dynamic evolution prediction of the machine tool in the next N control cycles. After the solution is completed at each time step, the vibration components of the spindle and feed axis motors are extracted from the obtained modal responses to obtain characteristic parameters such as the frequency, damping ratio, and response amplitude of each mode. After multi-step time-domain prediction, a continuous sequence of machine tool future state predictions is formed.

[0125] Based on the machine tool future state prediction sequence, a collaborative optimization objective function describing the comprehensive performance index is constructed; and the machine tool modal state vector is used as a constraint condition.

[0126] It should be noted that, based on the vibration response amplitude, phase shift, machining accuracy deviation, and energy consumption characteristics in the machine tool's future state prediction sequence, a collaborative optimization objective function reflecting the machine tool's comprehensive performance is defined. This collaborative optimization objective function comprehensively considers multiple indicators such as vibration suppression performance, drive efficiency, and trajectory accuracy. By using the modal parameters of each order in the machine tool's modal state vector as constraints, and introducing their frequency and damping ratio boundary ranges, the physical feasibility and stability of the machine tool's dynamic characteristics are ensured during the optimization solution process.

[0127] Based on the collaborative optimization objective function, a rolling optimization control strategy is used for real-time solution. In each optimization cycle, multiple time-domain steps are predicted to dynamically adjust the electrical drive control variables of the spindle-feed axis motor, so that the collaborative optimization objective function gradually converges and the optimal control parameter sequence is obtained.

[0128] Specifically, within each optimization cycle, using the machine tool modal state vector at the current moment as the initial condition, a future state prediction sequence for the machine tool over the next N control cycles (prediction time domain) is generated based on the virtual prediction model. A Receding Horizon Optimization (RHO) strategy is employed to formulate the optimization problem as finding the optimal control input sequence while satisfying both dynamic and state constraints. Among them, u t This represents the first control parameter, u. t+N-1 Let N represent the optimal control parameters at the final time step, and N represent the prediction time domain. The expression that minimizes the co-optimization objective function J is:

[0129]

[0130] in, It is the amplitude of the vibration response at time t. It is the trajectory tracking error at time t. α is the estimated driving energy consumption at time t; α is the weight of the vibration response amplitude, β is the weight of the trajectory tracking error, and γ is the weight of the estimated driving energy consumption. For example, the weights are allocated based on the priority requirements of vibration suppression, trajectory accuracy, and energy efficiency for specific processing technologies. In a finishing scenario where the primary goal is to achieve high-precision machining, β can be set to the highest, for example, 0.6; α to the next highest, for example, 0.3; and γ to a relatively low level, for example, 0.1, to ensure that resources are prioritized for ensuring machining accuracy.

[0131] The optimization process employs numerical optimization algorithms (such as Sequential Quadratic Programming (SQP) or the interior-point method) for online iterative computation. At each optimization step, a set of control parameter sequences for the future time domain is solved based on the current state prediction sequence and the co-optimization objective function. Subsequently, only the first control parameter u of this sequence is taken. t The servo drive at the current moment completes one optimization cycle. At the next sampling moment, starting from the new machine tool modal state vector, the prediction time domain is rolled, and the above optimization process is repeated to achieve dynamic adjustment of the electrical drive control variables of the spindle-feed axis motors and gradual convergence of the collaborative optimization objective function;

[0132] It should be noted that when the rate of change of the value of the collaborative optimization objective function is lower than the preset change threshold of 0.0002 within several consecutive control cycles, the optimization process is considered to have converged.

[0133] The optimal control parameter sequence is re-input into the virtual prediction model for dynamic simulation, and the vibration prediction results in the corresponding time domain are output.

[0134] Specifically, the optimal control parameter sequence is re-injected into the virtual prediction model as the electrical drive input, and multi-step forward simulation is run under the set dynamic boundary conditions and initial state variables. The vibration response results of the spindle-feed axis motor and structure under this control strategy are obtained by solving the machine tool dynamics equations in the time domain. Modal response features are extracted from the vibration prediction results to form the corresponding vibration prediction sequence in the time domain.

[0135] By introducing a virtual prediction model and a multi-step forward simulation mechanism, and combining the machine tool modal state vector to establish dynamic boundary constraints, real-time prediction of the machine tool's state and vibration response in the future time domain is achieved. A rolling optimization control strategy is employed to dynamically adjust the electrical drive parameters, enabling the control commands to be adaptive and forward-looking, thereby significantly improving vibration suppression accuracy and machine tool operational stability.

[0136] Methods for translating control parameter sequences into specific execution instructions and issuing them include:

[0137] Based on the optimal control parameter sequence, the electrical drive control variables corresponding to the spindle-feed axis motors for each time step are extracted; and according to the machine tool control interface protocol and servo drive logic, the mapping is translated into a set of control instructions that can be executed by the machine tool, and the corresponding execution instruction data packet is generated.

[0138] It should be noted that the setpoints of electrical drive control variables (such as current, voltage, or speed commands) need to be encapsulated according to the specific servo drive's communication protocol (e.g., EtherCAT, PROFIBUS, etc.). The translation process includes: quantizing continuous floating-point control parameters into integer data recognizable by the drive according to a preset scaling ratio and offset; and filling in the instruction code, target address, control mode, and parameter values ​​according to the frame format specified by the protocol (e.g., PDO mapping in the CiA 402 drive's specifications), and adding check bits (e.g., CRC check) to form a complete and error-free execution instruction data packet.

[0139] The execution instruction data packet is sent to the machine tool controller, which then drives the spindle-feed axis motor to perform the corresponding control.

[0140] It should be noted that the command delivery process utilizes a periodic synchronous communication mechanism via a real-time industrial Ethernet or fieldbus interface. The CNC system master station precisely sends the execution command data packets to the corresponding slave addresses of the spindle and each feed axis driver within a defined communication cycle. After receiving and parsing the data packets, the drivers update the setpoints of their internal control loops to the parameters specified by the command, thereby driving the motors to perform corresponding speed, torque, or position control.

[0141] Methods for real-time monitoring of the execution process and recording of actual response data include,

[0142] During the execution of instruction data packets, the machine tool operating status signals of the spindle-feed axis motors are monitored in real time to form the raw dataset for execution monitoring.

[0143] It should be noted that the monitoring process synchronously collects multi-source signals through a sensing system deployed on the machine tool. This includes: acquiring real-time drive signals of the motor through current and voltage sensors, acquiring actual position signals of the worktable or motor through grating rulers or encoders, and acquiring vibration signals of key points of the mechanical structure through accelerometers. These signals are collected synchronously at a uniform sampling frequency (10kHz for example) and accompanied by high-precision timestamps, collectively forming the raw dataset for execution monitoring containing electrical, positional, and vibration information.

[0144] Based on the original dataset of execution monitoring, the electrical drive characteristics and mechanical structure modal response are extracted, and the machine tool modal state vector during the control execution process of the computer tool is generated to generate the actual response data at the corresponding time-domain step.

[0145] It should be noted that the current and voltage signals in the original monitoring dataset undergo the same preprocessing and frequency domain transformation as before, extracting frequency domain features as the actual electrical drive characteristics. Simultaneously, the actual position signal is compared with the commanded position, and the trajectory deviation is calculated as the input for the actual mechanical structure modal response. Subsequently, this pair of actual features is input into the established "current-modal dynamic mapping relationship" to calculate the actual machine tool modal state vector reflecting the machine tool's true operating state. This actual machine tool modal state vector is continuously generated according to the control cycle (time domain step), forming an actual response data sequence corresponding to the predicted time domain.

[0146] Methods for calculating vibration response residuals based on actual response data and vibration prediction results include:

[0147] The actual response data and vibration prediction results are matched with the same time-domain step index, and the prediction-measurement comparison dataset is formed after aligning the timestamps.

[0148] It should be noted that the matching process is based on the timestamp of the control cycle. Since the vibration prediction result is obtained before the control command is issued, and the actual response data has response delay and sampling delay, a fixed time delay compensation (exemplarily 1-2 control cycles) needs to be introduced into the actual response data sequence to ensure that the predicted value and the measured value for the same control action are compared under the same time index, thereby forming a time-aligned prediction-measured comparison dataset.

[0149] For the machine tool modal state vector at the corresponding time step in the prediction-measurement comparison dataset, element-wise difference is performed to obtain the modal response difference sequence.

[0150] It should be noted that element-wise differencing refers to subtracting the corresponding element in the predicted machine tool modal state vector from the actual machine tool modal state vector at each aligned time point to obtain a difference vector. This difference vector contains the prediction errors of parameters such as modal frequencies, damping ratios, and response amplitudes at that moment. The difference vectors of all time-domain steps arranged in chronological order constitute the modal response difference sequence.

[0151] Based on the modal response difference sequence, the comprehensive difference index is calculated using the weighted mean square error criterion. The comprehensive difference index is defined as the vibration response residual corresponding to the time step.

[0152] It should be noted that the weighted mean square error criterion first assigns a weighting factor to the modal parameter error in the difference sequence; the weighting factor is set according to the contribution of each mode to the overall vibration energy or its influence on the processing quality (for example, the main vibration mode has a high weight, and the secondary modes have a low weight); the formula for calculating the comprehensive difference index is:

[0153]

[0154] Where ∈(t) is the comprehensive difference index, i.e., the vibration response residual, ΔY e (t) is the error component of the e-th modal parameter at time t, i.e., the difference ΔY in the machine tool modal state vector. t The corresponding e-th modal component, w e It is the weighting factor for the e-th mode; w e The settings are mainly based on the degree of influence of each mode on the total vibration energy or machining quality of the machine tool, and are usually determined by offline modal analysis or online energy calculation. For example, the main vibration mode with a large vibration contribution (such as the low-order mode with the largest amplitude) is given a higher weight (such as 0.3-0.5), while the secondary mode is given a lower weight (such as 0.05-0.1), and all weights are ensured to be 1.

[0155] Methods for online parameter correction based on the dynamic mapping relationship between current and mode and the virtual prediction model using vibration response residuals include:

[0156] Based on the vibration response residuals, the modal response difference sequences at each time step are analyzed to extract residual feature vectors that reflect the deviations in electrical drive characteristics and mechanical structure modal response. According to the contribution of each modal parameter in the machine tool modal state vector to the overall vibration energy, corresponding weights are set, and weighted processing is performed on the residual feature vectors to form a weighted residual feature set.

[0157] It should be noted that the analytical process first decomposes the modal response difference sequence into two parts according to its physical meaning: parameters directly related to the characteristics of the driving current and voltage spectrum (such as amplitude error in a specific frequency band) are classified as electrical drive characteristic deviations; parameters directly related to trajectory deviations and vibration amplitudes (such as modal damping and amplitude errors) are classified as mechanical structure modal response deviations. These electrical drive characteristic deviations and mechanical structure modal response deviations are then concatenated in sequence to form the residual feature vector.

[0158] Based on the contribution ratio of the variance of each modal parameter in the machine tool modal state vector to the total vibration energy, an appropriate weight is assigned to each element in the residual feature vector (for example, the weight of the main vibration modal parameter is 0.3, and the weights of other modes decrease sequentially). After weighting all elements, a weighted residual feature set is formed.

[0159] The weighted residual feature set is input into the current-modal dynamic mapping relationship. Based on the residual feature vector, the associated parameters are recursively adjusted so that the dynamic influence matrix between the electrical drive characteristics and the mechanical structure modal response reconverges in the time series, resulting in the corrected current-modal dynamic mapping relationship.

[0160] It should be noted that the adjustment process employs a recursive least squares method with a forgetting factor. The weighted residual feature set is used as a new observation error and input along with the current electrical drive characteristics into the parameter update algorithm for the current-modal dynamic mapping relationship. Based on the magnitude and direction of the residuals, the corresponding elements in the dynamic influence matrix are automatically adjusted; for example, the correction amount of the associated parameters corresponding to the drive frequency bands that generate large residuals is increased. Through multiple recursive updates, the error between the predicted output of the current-modal dynamic mapping relationship and the actual response data sequence gradually decreases until the changes in the associated parameters tend to stabilize, thereby obtaining a corrected current-modal dynamic mapping relationship that better matches the current actual operating conditions.

[0161] Based on the corrected current-modal dynamic mapping relationship and vibration response residuals, the machine tool modal state vector in the virtual prediction model is re-estimated; and the parameters of the machine tool dynamic equation in the virtual prediction model are updated with the new machine tool modal state vector.

[0162] It should be noted that, using the corrected current-modal dynamic mapping relationship and combining it with the latest acquired electrical drive characteristics, the machine tool modal state vector at the current moment is recalculated as a more accurate re-estimate. Then, this re-estimated modal state vector (including updated natural frequencies, damping ratios, and other parameters) is substituted into the dynamic equations of the virtual prediction model, specifically updating the corresponding parameters of the modal mass, modal stiffness, or modal damping matrix in the equations. For example, the corresponding modal stiffness coefficient is fine-tuned based on a re-estimated frequency. In this way, the internal parameters of the virtual prediction model are updated online, making it more accurately reflect the current dynamic characteristics of the machine tool structure, thereby improving the accuracy of the prediction for the next cycle.

[0163] This embodiment also provides an adaptive control system for vibration suppression of CNC machine tools, including:

[0164] The feature extraction module is used to extract frequency and time domain features from the multi-source sensing signals of the spindle-feed axis motor and generate vibration state feature vectors.

[0165] The mapping modeling module is used to establish the current-modal dynamic mapping relationship between electrical drive characteristics and mechanical structure modal response based on vibration state feature vectors; and to obtain real-time machine tool modal state vectors through online recursive estimation.

[0166] The prediction and optimization module is used to inject the machine tool modal state vector as boundary conditions into the virtual prediction model; and through the rolling optimization control strategy with the goal of multi-performance index synergistic optimization, it solves to obtain the optimal control parameter sequence and corresponding vibration prediction results in the future time domain.

[0167] The execution monitoring module is used to translate the control parameter sequence into specific execution instructions and issue them; monitor the execution process in real time and record actual response data;

[0168] The online correction module is used to calculate the vibration response residual based on the actual response data and vibration prediction results; and to perform online parameter correction on the current-mode dynamic mapping relationship and the virtual prediction model based on the vibration response residual.

[0169] This embodiment also provides a computer device applicable to the adaptive control method for vibration suppression of CNC machine tools, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the adaptive control method for vibration suppression of CNC machine tools as proposed in the above embodiment.

[0170] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0171] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements the adaptive control method for vibration suppression of CNC machine tools as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0172] In summary, this invention achieves real-time coupled modeling of electrical drive characteristics and mechanical structure modes by establishing a dynamic mapping relationship between the electromagnetic drive characteristics of the servo spindle and feed motor and the dynamic response of the machine tool structure. Based on this mapping relationship, an adaptive vibration control model is constructed. By introducing a dynamic response feature recognition and optimal control parameter self-learning mechanism, control commands can be updated in real time during machine tool operation, allowing the control strategy to dynamically adjust with changes in operating state, thus overcoming the problems of control lag and static modeling in traditional methods. This technology not only improves the accuracy of vibration prediction and the real-time performance of control response but also enhances system stability and machining accuracy under complex conditions such as high-speed cutting and multi-axis linkage, achieving active suppression of machine tool vibration and adaptive performance optimization.

[0173] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. An adaptive control method for vibration suppression in CNC machine tools, characterized in that: include, Frequency and time domain features are extracted from the multi-source sensor signals of the spindle-feed axis motor to generate a vibration state feature vector; Based on the vibration state feature vector, a current-modal dynamic mapping relationship between electrical drive characteristics and mechanical structure modal response is established; and the real-time machine tool modal state vector is obtained through online recursive estimation. The machine tool modal state vector is injected as a boundary condition into the virtual prediction model; and the optimal control parameter sequence and corresponding vibration prediction results in the future time domain are obtained by using a rolling optimization control strategy with the goal of synergistic optimization of multiple performance indicators. Translate the control parameter sequence into specific execution instructions and issue them; Real-time monitoring of the execution process and recording of actual response data; The vibration response residual is calculated based on the actual response data and vibration prediction results; the current-mode dynamic mapping relationship and the virtual prediction model are then corrected online based on the vibration response residual.

2. The adaptive control method for vibration suppression of CNC machine tools as described in claim 1, characterized in that: The method for generating vibration state feature vectors includes, For the multi-source sensor signals of the spindle-feed axis motor, mean subtraction processing to eliminate DC offset and low-pass filtering processing to suppress spectral aliasing are performed. A fixed-duration sliding window is used to continuously segment the preprocessed multi-source sensor signal; each segmented multi-source sensor signal is converted into a corresponding frequency domain amplitude spectrum to extract frequency domain features; and the position feedback signal of the feed motor is compared with the command position signal output by the servo controller to obtain the trajectory deviation and extract time domain features. The frequency domain features and time domain features are concatenated and combined, and the values ​​are normalized to generate a vibration state feature vector.

3. The adaptive control method for vibration suppression of CNC machine tools as described in claim 2, characterized in that: The method for establishing the current-mode dynamic mapping relationship includes, Frequency domain features are separated from the vibration state feature vector, and frequency domain response features reflecting electromagnetic drive characteristics are extracted to form electrical drive characteristics; time domain features are separated from the vibration state feature vector, and time domain response features reflecting mechanical dynamic characteristics are extracted to form mechanical structure modal response. Time alignment and synchronization pairing are performed on the electrical drive characteristics and mechanical structure modal response to form an electrical-mechanical corresponding time series dataset; Based on the variation patterns of the electrical-mechanical time series dataset, a multivariate correlation relationship is established with electrical drive characteristics as input and mechanical structure modal response as output. The correlation weights between the electrical drive characteristics and the modal response of the mechanical structure under different time delays are calculated to construct a dynamic influence matrix; The adjustable parameters in the dynamic influence matrix used to characterize the intensity and direction of the influence of electrical drive characteristics on the modal response of mechanical structures are defined as correlation parameters. The correlation parameters are adaptively adjusted by a recursive update method so that the multivariate correlation relationship continues to converge in the time series, thereby establishing a current-modal dynamic mapping relationship.

4. The adaptive control method for vibration suppression of CNC machine tools as described in claim 3, characterized in that: The method for obtaining real-time machine tool modal state vectors through online recursive estimation includes: Based on the current-mode dynamic mapping relationship, the electrical drive characteristics and mechanical structure modal response at the current moment are received, and synchronous pairing is performed according to the timestamp to construct an electrical-mechanical timing input set; Based on the electrical-mechanical timing input set and the current-modal dynamic mapping relationship, the estimated values ​​of the machine tool modal parameters at the current moment are calculated, and the initial machine tool modal state vector is generated. The initial machine tool modal state vector is compared with the machine tool modal state vector at the previous moment. The correlation parameters are dynamically adjusted according to the differences to correct the mapping accuracy and obtain the corrected current-modal dynamic mapping relationship. Based on the corrected current-modal dynamic mapping relationship and the electrical-mechanical timing input set, online recursive estimation calculation is performed to obtain the machine tool modal state vector in real time.

5. The adaptive control method for vibration suppression of CNC machine tools as described in claim 4, characterized in that: The method for obtaining the optimal control parameter sequence and corresponding vibration prediction results in the future time domain includes: Input the machine tool modal state vector into the virtual prediction model; determine the dynamic boundary conditions of the machine tool based on the structural modal characteristic parameters in the machine tool modal state vector; assign the dynamic boundary conditions to the initial state variables of the virtual prediction model; Based on a multi-step forward simulation architecture of virtual prediction model, the current control command and historical operation data are integrated in real time. Through time-domain iterative solution of machine tool dynamic equations, the state evolution trajectory within N control cycles is deduced under kinematic and dynamic constraints. Simultaneously, the coupled vibration modal response of the spindle-feed axis motor is calculated to form a prediction sequence of the future state of the machine tool. Based on the machine tool future state prediction sequence, a collaborative optimization objective function describing the comprehensive performance index is constructed; and the machine tool modal state vector is used as a constraint condition. Based on the collaborative optimization objective function, a rolling optimization control strategy is adopted for real-time solution. In each optimization cycle, multiple time-domain steps are predicted by sliding, and the electrical drive control variables of the spindle-feed axis motor are dynamically adjusted so that the collaborative optimization objective function gradually converges and the optimal control parameter sequence is obtained. The optimal control parameter sequence is re-input into the virtual prediction model for dynamic simulation, and the vibration prediction results in the corresponding time domain are output.

6. The adaptive control method for vibration suppression of CNC machine tools as described in claim 5, characterized in that: The method for translating the control parameter sequence into specific execution instructions and issuing them includes: Based on the optimal control parameter sequence, the electrical drive control variables corresponding to the spindle-feed axis motors for each time step are extracted; Based on the machine tool control interface protocol and servo drive logic, the mapping is translated into a set of control instructions that the machine tool can execute, and the corresponding execution instruction data packet is generated; The execution instruction data packet is sent to the machine tool controller, which then drives the spindle-feed axis motor to perform the corresponding control.

7. The adaptive control method for vibration suppression of CNC machine tools as described in claim 6, characterized in that: The method for real-time monitoring of the execution process and recording actual response data includes, During the execution of instruction data packets, the machine tool operating status signals of the spindle-feed axis motors are monitored in real time to form the raw dataset of execution monitoring; Based on the original dataset of execution monitoring, the electrical drive characteristics and mechanical structure modal response are extracted, and the machine tool modal state vector during the control execution process of the computer tool is generated to generate the actual response data at the corresponding time-domain step.

8. The adaptive control method for vibration suppression of CNC machine tools as described in claim 7, characterized in that: The method for calculating the vibration response residual based on actual response data and vibration prediction results includes: The actual response data and vibration prediction results are matched with the same time-domain step index, and the prediction-measurement comparison dataset is formed after aligning the timestamps. For the machine tool modal state vector at the corresponding time step in the prediction-measurement comparison dataset, perform element-wise difference to obtain the modal response difference sequence; Based on the modal response difference sequence, the comprehensive difference index is calculated using the weighted mean square error criterion. The comprehensive difference index is defined as the vibration response residual corresponding to the time step.

9. The adaptive control method for vibration suppression of CNC machine tools as described in claim 8, characterized in that: The method for online parameter correction of the current-modal dynamic mapping relationship and the virtual prediction model based on the vibration response residual includes: Based on the vibration response residuals, the modal response difference sequences at each time step are analyzed to extract residual feature vectors that reflect the deviations of electrical drive characteristics and mechanical structure modal response. According to the contribution of each modal parameter in the machine tool modal state vector to the overall vibration energy, corresponding weights are set, and weighted processing is performed on the residual feature vectors to form a weighted residual feature set. The weighted residual feature set is input into the current-modal dynamic mapping relationship. Based on the residual feature vector, the associated parameters are recursively adjusted so that the dynamic influence matrix between the electrical drive characteristics and the mechanical structure modal response reconverges in the time series, resulting in the corrected current-modal dynamic mapping relationship. Based on the corrected current-modal dynamic mapping relationship and vibration response residuals, the machine tool modal state vector in the virtual prediction model is re-estimated; and the parameters of the machine tool dynamic equation in the virtual prediction model are updated with the new machine tool modal state vector.

10. A CNC machine tool vibration suppression adaptive control system, based on the CNC machine tool vibration suppression adaptive control method according to any one of claims 1 to 9, characterized in that: include, The feature extraction module is used to extract frequency and time domain features from the multi-source sensing signals of the spindle-feed axis motor and generate vibration state feature vectors. The mapping modeling module is used to establish the current-modal dynamic mapping relationship between electrical drive characteristics and mechanical structure modal response based on vibration state feature vectors; and to obtain real-time machine tool modal state vectors through online recursive estimation. The prediction and optimization module is used to inject the machine tool modal state vector as boundary conditions into the virtual prediction model; and through the rolling optimization control strategy with the goal of multi-performance index synergistic optimization, it solves to obtain the optimal control parameter sequence and corresponding vibration prediction results in the future time domain. The execution monitoring module is used to translate the control parameter sequence into specific execution instructions and issue them. Real-time monitoring of the execution process and recording of actual response data; The online correction module is used to calculate the vibration response residual based on the actual response data and vibration prediction results; and to perform online parameter correction on the current-mode dynamic mapping relationship and the virtual prediction model based on the vibration response residual.

Images