Spindle vibration active suppression and control system

Abstract

The application discloses a spindle vibration active suppression and control system and belongs to the technical field of spindle vibration control. The system comprises a vibration acquisition and judgment module, which acquires multi-dimensional vibration signals of a spindle in a running state, carries out time domain and frequency domain feature separation operations, and judges whether the current vibration state of the spindle reaches a preset active suppression intervention condition according to a separation result; a feature space module, which carries out nonlinear dynamics feature extraction on the multi-dimensional vibration signals when the active suppression intervention condition is reached, and constructs a vibration feature space of the spindle according to the distribution law of the extracted nonlinear dynamics features in a reconstructed phase space; and an execution closed loop module, which drives an active actuator to output a reverse suppression force according to a feedforward control instruction, synchronously acquires residual vibration signals, carries out closed loop correction on the feedforward control instruction, and generates a closed loop control signal.

Description

Technical Field

[0001] This invention relates to the field of spindle vibration control technology, specifically to an active suppression and control system for spindle vibration. Background Technology

[0002] As a core component of rotating machinery, the vibration generated by the spindle during operation directly affects machining accuracy, surface quality, and equipment lifespan. Existing spindle vibration control technologies mostly employ passive dampers or active control methods based on frequency domain filtering. Passive dampers absorb vibration energy in specific frequency bands through elastic elements or damping materials, and their control parameters are fixed after design and installation, making it difficult to adapt to vibration characteristic drift caused by changes in spindle speed, load, and other operating conditions. Active control methods based on frequency domain filtering rely on techniques such as Fast Fourier Transform (FFT) to obtain vibration spectrum information, thereby generating suppression signals. These methods treat the vibration signal as a linear stationary process, ignoring the nonlinear dynamic behavior exhibited by the spindle under high speed and variable load conditions. When the spindle system exhibits complex vibration patterns such as chaos or bifurcation, linear analysis methods cannot effectively extract features reflecting the system's inherent dynamic evolution, leading to a mismatch between the generated suppression signal and the actual vibration state. The suppression effect rapidly diminishes as nonlinearity increases. Furthermore, conventional active control schemes often use lookup tables or fixed control laws to generate control parameters, resulting in parameter updates lagging behind the instantaneous changes in vibration. When the spindle starts, stops, changes speed, or encounters sudden disturbances, the vibration amplitude and phase change drastically within a millisecond timescale. Fixed control parameters cannot keep up with this rapid dynamic, causing vibration energy to be transferred to the structure in a short period of time, forming impact loads. The problem this invention aims to solve is how to establish a vibration characterization method that reflects the true dynamic integrity of the system after the spindle vibration enters the strongly nonlinear region, and how to match the initial control parameters accordingly; and how to keep the adjustment rate of the control parameters synchronized with the dynamic evolution of the vibration when the vibration changes drastically in an instant. Summary of the Invention

[0003] The purpose of this invention is to provide an active suppression and control system for spindle vibration, which aims to solve the problem of control parameter matching deviation caused by the inability of the vibration characterization method to reflect the true dynamic integrity of the system after the spindle vibration enters the strong nonlinear region. At the same time, it also solves the problem that the control parameter adjustment rate cannot keep pace with the instantaneous dynamic evolution of vibration.

[0004] To achieve the above objectives, the present invention provides the following technical solution: The present invention provides a spindle vibration active suppression and control system, which includes a vibration acquisition and judgment module, a feature space module, a parameter adaptation module, a feedforward correction module, and an execution closed-loop module. The vibration acquisition and judgment module acquires multi-dimensional vibration signals of the spindle in operation and performs time-domain and frequency-domain feature separation operations on the signals. Based on the separation results, it determines whether the current vibration state of the spindle meets the preset active suppression intervention conditions. This module effectively separates high-frequency and low-frequency fluctuation components in the signal and uses the ratio of high-frequency energy to low-frequency energy in the overall vibration as the judgment basis. It can sensitively capture early high-frequency vibration signs caused by abnormal factors such as cutting chatter and bearing damage, avoiding unnecessary active suppression actions under stable operating conditions, thereby balancing system response speed and energy consumption.

[0005] As a technical solution of this invention, when the vibration acquisition and judgment module performs feature separation operations in the time and frequency domains, it performs empirical mode decomposition on the multi-dimensional vibration signal to obtain multiple intrinsic mode components, which are then divided into a high-frequency component set and a low-frequency component set according to the order of frequency from high to low. The instantaneous average energy of each intrinsic mode component in the high-frequency component set and the instantaneous average energy of each intrinsic mode component in the low-frequency component set are calculated respectively. The ratio of the high-frequency instantaneous energy average to the low-frequency instantaneous energy average is used as the high-frequency proportion coefficient of vibration energy. When this high-frequency proportion coefficient of vibration energy is greater than a preset high-frequency energy proportion threshold, it is determined that the spindle has reached the active suppression intervention condition. This method performs adaptive decomposition based on the scale characteristics of the signal itself, without the need for preset basis functions. It has good adaptability to non-stationary vibration signals under varying spindle speeds and loads, and the high-frequency energy proportion coefficient can intuitively reflect the sudden increase of micro-amplitude high-frequency flutter components, effectively improving the timeliness and accuracy of intervention judgment.

[0006] When the active suppression intervention condition is met, the feature space module extracts nonlinear dynamic features from the multi-dimensional vibration signal and constructs the vibration feature space of the principal shaft based on the distribution law of the extracted nonlinear dynamic features in the reconstructed phase space. Principal shaft vibration inherently exhibits nonlinear dynamic behavior. This module mines the implicit dynamic attractor features from the temporal evolution structure of the vibration signal and maps them to a feature space that reflects the intrinsic state changes of the system. This provides a deeper state description for subsequent control parameter decisions, overcoming the problem that traditional time-frequency features cannot adequately distinguish between different vibration sources and different instability stages.

[0007] Preferably, the process of nonlinear dynamic feature extraction in the feature space module includes: for each dimension of the vibration signal, determining the time delay parameter using the mutual information method and determining the embedding dimension parameter using the pseudo-nearest neighbor method; reconstructing the phase space of the vibration signal based on the time delay parameter and the embedding dimension parameter to obtain the reconstructed trajectory matrix corresponding to each dimension; calculating the Euclidean distance between all adjacent trajectory points in the reconstructed trajectory matrix and using the mean of the Euclidean distances of all adjacent trajectory points as the average trajectory distance for that dimension; constructing vibration space feature vectors according to the dimension numbering order of all dimensions and mapping them to a preset polar coordinate system to obtain the vibration feature space. Selecting the time delay using the mutual information method can better preserve the nonlinear correlation of the signal, while the embedding dimension parameter determined by the pseudo-nearest neighbor method ensures that the phase space is fully expanded without generating false neighbor points. The constructed vibration feature space integrates multi-dimensional vibration information into a unified geometric description, facilitating subsequent matching and discrimination with different vibration modes.

[0008] The parameter adaptation module dynamically matches an initial control parameter combination corresponding to the current vibration state from a control strategy library based on the projection intensity of the vibration feature space onto multiple preset vibration mode categories. As a specific implementation, the module calculates the spatial distance between the center point of the vibration feature space and the center points of each of the multiple preset vibration mode categories, and selects the three vibration mode categories with the smallest spatial distances as candidate vibration mode categories. It then obtains the corresponding control strategy nodes for each candidate vibration mode category in the control strategy library, extracts stiffness adjustment parameters and damping adjustment parameters from each control strategy node, and weights and fuses all extracted stiffness and damping adjustment parameters according to the reciprocal of their corresponding spatial distances to obtain the initial control parameter combination. In the weighted fusion, the weight of each stiffness and damping adjustment parameter is set to the ratio of the reciprocal of its corresponding spatial distance to the sum of the reciprocals of all the aforementioned spatial distances. This method uses the proximity of the vibration feature space to the known vibration mode category center to represent the membership degree of the current state with each mode. By fusing the control parameters of multiple nearest neighbor modes through reciprocal weighting, it avoids parameter mutations caused by hard matching of a single mode and can output smooth and appropriate initial control parameters for mode aliasing that may occur during the actual vibration of the spindle, thus enhancing the robustness and accuracy of parameter adaptation.

[0009] The feedforward correction module dynamically corrects the initial control parameter combination based on the instantaneous change trend of the multi-dimensional vibration signal, generating feedforward control commands. During the feedforward correction process, it acquires the instantaneous amplitude sequence of the multi-dimensional vibration signal from multiple consecutive historical moments prior to the current moment and performs polynomial fitting to obtain the instantaneous amplitude change curve. It then calculates the first and second derivatives of this instantaneous amplitude change curve at the current moment, determining the direction of vibration acceleration change of the principal shaft based on these values. If the direction of vibration acceleration change is an accelerating divergent direction, the stiffness and damping adjustment parameters in the initial control parameter combination are increased by a preset first step size to obtain the feedforward control command. If the direction of vibration acceleration change is a decelerating convergent direction, the stiffness and damping adjustment parameters in the initial control parameter combination are decreased by a preset second step size to obtain the feedforward control command. This feedforward correction mechanism anticipates the development trend of vibration based on the instantaneous trend of vibration amplitude changes. When accelerating and diverging, it strengthens the suppression force to block the malignant growth of amplitude. When decelerating and converging, it appropriately reduces the suppression intensity to avoid system stiffening or energy waste caused by over-control, thereby achieving real-time tracking and reasonable adjustment of control quantity and vibration trend.

[0010] The closed-loop execution module drives the active actuator on the spindle to output a reverse damping force according to the feedforward control command, and simultaneously acquires the residual vibration signal after the active actuator outputs the reverse damping force. Based on the residual vibration signal, the feedforward control command is corrected in a closed-loop manner to generate a closed-loop control signal. In the closed-loop correction, the residual amplitude of the residual vibration signal at multiple preset frequency points is acquired, and these residual amplitudes are constructed into a residual amplitude vector in ascending order of frequency. The residual amplitude vector is compared element-by-element with a preset residual amplitude tolerance vector to determine if there are any frequency points in the residual amplitude vector that exceed the corresponding element in the residual amplitude tolerance vector. If such frequency points exist, the excess amplitude difference corresponding to all excess frequency points is acquired, and the sum of all excess amplitude differences is taken as the residual vibration excess value. A correction coefficient is generated based on the residual vibration excess value, and this correction coefficient is multiplied by the stiffness adjustment parameter and damping adjustment parameter in the feedforward control command to obtain the closed-loop control signal. This closed-loop correction method directly performs a fine evaluation of the frequency components of residual vibration and compensates for specific frequency components that exceed the tolerance limit. It can effectively eliminate the narrowband vibration peaks remaining after feedforward control and further improve the vibration suppression effect of the spindle across the entire frequency band.

[0011] Preferably, the residual amplitude tolerance vector is preset based on the fundamental vibration amplitude spectrum of the spindle at the current speed, and this fundamental vibration amplitude spectrum is updated each time the spindle is started. This allows the tolerance threshold to adapt to the inherent vibration characteristics of the spindle at different speeds and the normal baseline drift caused by gradually changing factors such as wear and temperature, reducing the occurrence of miscorrection and undercorrection.

[0012] The active vibration suppression and control system of this invention starts with the separation and intervention determination of multi-dimensional features of vibration signals. It constructs a state space characterizing the essence of vibration through nonlinear dynamic feature extraction and phase space reconstruction. Weighted matching of the vibration feature space and a mode library yields initial control parameters that closely match the actual vibration state. Then, feedforward correction is performed using the instantaneous change trend of vibration to enhance the timeliness of control. Finally, precise force compensation output is achieved by combining frequency-based closed-loop correction of the residual vibration signal. The entire system integrates deep perception of vibration state, adaptive parameter decision-making, and coordinated adjustment of feedforward and closed-loop, significantly improving the suppression performance of wideband vibration, transient chatter, and multimodal composite vibration of the spindle, ensuring machining accuracy and the reliability of the spindle system operation.

[0013] The technical effects and advantages provided by the present invention in the above technical solution are as follows: When extracting nonlinear dynamic features from multidimensional vibration signals, the mutual information method and the pseudo-nearest neighbor method are used to determine the time delay parameters and embedding dimension parameters of each dimension of the vibration signal, respectively. Based on this, phase space reconstruction is performed on the vibration signal of each dimension to obtain its respective reconstructed trajectory matrix. In the reconstructed phase space, the nonlinear structure of the vibration signal is fully unfolded in the sense of topological equivalence. The high-dimensional dynamic information originally implicit in the univariate time series is mapped to the distribution and evolution of trajectory points in the phase space. The average Euclidean distance between adjacent trajectory points in the reconstructed trajectory matrix is ​​calculated and used as an element of the vibration space feature vector, so that the nonlinear characteristics of the principal axis vibration can be quantitatively characterized. The vibration space feature vector is further mapped to a polar coordinate system to form a vibration feature space. The coordinate reference built into this space makes the feature vectors corresponding to different vibration modes naturally comparable in terms of angle and radial distance. When initial control parameters are matched based on the projection intensity of the vibration characteristic space onto a preset vibration mode category, the similarity between vibration states is measured by spatial distance rather than a simple amplitude threshold. Complex vibration patterns generated by the main shaft in the strongly nonlinear region can be accurately classified, improving the fit between the matched control parameter combination and the actual vibration state, thus reducing control deviations caused by mode misjudgment at the source. After obtaining the initial control parameter combination, instantaneous amplitude values ​​at multiple historical moments are continuously collected, and an instantaneous amplitude change curve is fitted accordingly. The first and second derivatives of this curve at the current moment are calculated to determine the direction of the main shaft's vibration acceleration change. Instantaneous stiffness and damping adjustments are made directionally based on whether the vibration is trending towards accelerated divergence or decelerated convergence. When the direction of vibration acceleration indicates accelerated divergence, the stiffness and damping adjustment parameters are increased by a preset first step size to actively inject more constraint energy into the system to curb vibration propagation. When the direction indicates decelerated convergence, the stiffness and damping adjustment parameters are decreased by a preset second step size to avoid unnecessary energy accumulation in the structure due to excessive constraint. The real-time updates of stiffness and damping parameters following the vibration trend enable the generation of feedforward control commands to be synchronized with the instantaneous changes in vibration signals. This fills the response gap caused by traditional fixed-parameter control laws when facing rapid dynamic changes in vibration, and dynamically suppresses vibration impacts caused by transient conditions such as start-stop and speed change. Attached Figure Description

[0014] 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.

[0015] Figure 1 This is a schematic diagram of the active suppression and control system for spindle vibration; Figure 2 This is a flowchart of the vibration sampling and judgment module; Figure 3 This is a flowchart of the feature space module processing; Figure 4 This is a flowchart of the control parameter generation process for the parameter adaptation module; Figure 5 This is a flowchart of the feedforward correction module. Figure 6 This is the closed-loop control flowchart for executing the closed-loop module. Detailed Implementation

[0016] 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.

[0017] See Figure 1 This invention provides a spindle vibration active suppression and control system, which includes a vibration acquisition and judgment module, a feature space module, a parameter adaptation module, a feedforward correction module, and an execution closed-loop module. The vibration acquisition and judgment module acquires multi-dimensional vibration signals of the spindle during operation, performs time-domain and frequency-domain feature separation, and determines whether the current vibration state of the spindle meets preset active suppression intervention conditions based on the separation results. When the active suppression intervention conditions are met, the feature space module extracts nonlinear dynamic features from the multi-dimensional vibration signals and constructs a vibration feature space of the spindle based on the distribution law of the extracted nonlinear dynamic features in the reconstructed phase space. The parameter adaptation module dynamically matches an initial control parameter combination corresponding to the current vibration state from a control strategy library based on the projection intensity of the vibration feature space onto multiple preset vibration mode categories. The feedforward correction module dynamically corrects the initial control parameter combination based on the instantaneous change trend of the multi-dimensional vibration signals, generating feedforward control commands. The execution closed-loop module drives the active actuator on the spindle to output a reverse damping force according to the feedforward control command, and simultaneously collects the residual vibration signal after the active actuator outputs the reverse damping force. Then, it performs closed-loop correction on the feedforward control command based on the residual vibration signal to generate a closed-loop control signal.

[0018] Example 1: In specific implementation, please refer to Figure 2The vibration acquisition and judgment module acquires multi-dimensional vibration signals of the spindle in operation through multiple vibration sensors arranged on the spindle bearing housing and spindle housing. The multi-dimensional vibration signals include at least vibration acceleration signals in two mutually perpendicular radial directions of the spindle and vibration displacement signals in the axial direction.

[0019] The vibration acquisition and judgment module performs empirical mode decomposition (EMD) on each dimension of the multi-dimensional vibration signal. The EMD process is as follows: identify all local maxima and local minima of each dimension of the vibration signal; fit the upper and lower envelopes using cubic spline interpolation; calculate the mean curves of the upper and lower envelopes; subtract the mean curves from the original vibration signal to obtain the intermediate signal; determine whether the intermediate signal meets the conditions for intrinsic modal components (IMCs). The conditions for IMCs are that the number of extreme points and the number of zero-crossing points are equal or differ by one over the entire time range, and the mean of the upper and lower envelopes is zero at any time; if the intermediate signal does not meet the conditions for IMCs, repeat the above steps with the intermediate signal as a new input until the first IMC that meets the conditions is obtained; separate the first IMC from the original vibration signal to obtain the residual signal; repeat the entire decomposition process on the residual signal to obtain the second IMC, the third IMC, and so on until the residual signal becomes a monotonic function that cannot be further decomposed. Therefore, each dimensional vibration signal is decomposed into several intrinsic modal components, and the number of all intrinsic modal components is denoted as . .

[0020] The vibration acquisition and judgment module obtains the vibration signal in each dimension. Each intrinsic modal component (IMC) is analyzed using a Fast Fourier Transform (FFT) to calculate its center frequency. The center frequency is determined by taking the frequency value corresponding to the highest amplitude point in the IMC spectrum. All IMCs in each dimension of the vibration signal are sorted in descending order of their center frequencies, resulting in a sorted IMC sequence. The vibration sampling and analysis module obtains the average center frequency of all IMCs. IMCs with center frequencies higher than the average are grouped into a high-frequency component set, while those with center frequencies lower than or equal to the average are grouped into a low-frequency component set. For multi-dimensional vibration signals, the high-frequency component sets from all dimensions are merged into a final high-frequency component set, and the low-frequency component sets from all dimensions are merged into a final low-frequency component set.

[0021] The vibration acquisition and analysis module calculates the instantaneous energy value for each eigenmode component in the high-frequency component set using the Hilbert transform. For a single eigenmode component... Its Hilbert transformation is Constructing analytic signals ,in, If the imaginary unit is used, then the instantaneous amplitude of this intrinsic mode component is... The instantaneous energy value is calculated using the following formula: in, Indicates the index in the high-frequency component set. The intrinsic mode components at time The instantaneous energy value, For the intrinsic mode components at time... instantaneous amplitude, Sampling time, This refers to the index number of the intrinsic mode component in the high-frequency component set. , This represents the total number of intrinsic mode components in the high-frequency component set.

[0022] The vibration sampling and analysis module calculates the average instantaneous energy value of each intrinsic modal component in the high-frequency component set at all sampling times, obtaining the average instantaneous energy value of each intrinsic modal component. Then, it calculates the average instantaneous energy value of all intrinsic modal components in the high-frequency component set, and uses this average as the high-frequency instantaneous energy mean. For each intrinsic modal component in the low-frequency component set, the vibration sampling and analysis module uses the same instantaneous energy value calculation method and average instantaneous energy value calculation method as for the high-frequency component set, obtaining the average instantaneous energy value of each intrinsic modal component. Then, it calculates the average instantaneous energy value of all intrinsic modal components in the low-frequency component set, and uses this average as the low-frequency instantaneous energy mean.

[0023] The vibration acquisition and analysis module uses the ratio of the average high-frequency instantaneous energy to the average low-frequency instantaneous energy as the high-frequency proportion coefficient of vibration energy. The high-frequency energy proportion threshold is pre-stored in the vibration acquisition and analysis module. The high-frequency energy proportion threshold is determined as follows: under known stable operating conditions with no abnormal vibrations, multiple sets of multi-dimensional vibration signals are acquired. The high-frequency proportion coefficient of vibration energy corresponding to each set of multi-dimensional vibration signals is calculated using the same steps described above, resulting in multiple samples of high-frequency proportion coefficients. The mean of these multiple samples is then calculated. and standard deviation Set the high-frequency energy percentage threshold to .

[0024] The vibration sampling and judgment module compares the high-frequency proportion coefficient of the vibration energy currently calculated with the high-frequency energy proportion threshold. If the high-frequency proportion coefficient of the vibration energy is greater than the high-frequency energy proportion threshold, it determines that the main shaft has reached the active suppression intervention condition and sends a start command to the feature space module to trigger the subsequent vibration feature space construction process.

[0025] Example 2: In specific implementation, please refer to Figure 3 The feature space module receives a trigger signal from the vibration acquisition and judgment module and performs nonlinear dynamic feature extraction on each dimension of the multi-dimensional vibration signal. The number of dimensions of the multi-dimensional vibration signal is denoted as... The vibration signal in each dimension is a discrete-time series, where the th... The dimensional vibration signal is denoted as , , , The total number of sampling points for the vibration signal in each dimension.

[0026] For the 3D vibration signal The feature space module uses the mutual information method to determine the time delay parameter. The implementation process of the mutual information method is as follows: the vibration signal... The amplitude range is divided into Statistical analysis of vibration signals in equidistant intervals. probability distribution within each interval , For a given time delay Constructing a delayed signal Statistical vibration signals and delayed signal joint probability distribution , , Calculate the mutual information value using the following formula. : in, Indicates the first 3D vibration signal with time delay The mutual information value below, Indicates the first The amplitude of the dimensional vibration signal falls into the 1st... The probability of each interval. Indicates the first The amplitude of the delayed signal of the dimensional vibration signal falls into the 3rd generation. The probability of each interval. Indicates the first The amplitude of the dimensional vibration signal falls into the 1st... The interval and the amplitude of the delayed signal falls within the first interval. The joint probability of each interval. This represents the total number of intervals into which the amplitude range is equally divided. The value of is an integer determined through stability analysis of variance estimation, specifically based on the fact that... Starting from 10 and increasing in steps of 5 up to 100, select a mutual information function. The minimum value corresponding to the first time the standard deviation of the curve fluctuation falls below the preset fluctuation threshold of 0.05. value.

[0027] Feature space module delays time Starting from 1 and incrementing by 1, calculate each... Mutual information value corresponding to the value To obtain the mutual information function Follow The changing curve. The feature space module scans the mutual information function curve, identifying all local minima on the curve. The criterion for determining a local minimum is that the mutual information value at that point is less than the mutual information values ​​of the two adjacent points. The feature space module records the time delay corresponding to the first local minimum reached by the mutual information function, and uses this time delay as the... The time delay parameter of the vibration signal is denoted as . .

[0028] In determining the first Time delay parameters of vibration signal Next, the feature space module uses the pseudo-nearest neighbor method to determine the embedding dimension parameters. The implementation process of the pseudo-nearest neighbor method is as follows: setting trial values ​​for the embedding dimension. ,make Start from 1 and increment by 1 step; for each trial value Using the determined time delay parameter , will the 3D vibration signal Reconstructed A sequence of phase points in a 3D phase space, where the th phase point is... The vectors of the phase points are: , , ;exist In the 3D phase space, for each phase point Find the nearest neighboring phase point with the Euclidean distance. , ; Phase point and neighboring phase points Embedded into In the phase space, we obtain and ;calculate and Euclidean distance between them and If the ratio of the Euclidean distance between them is greater than a preset false neighbor determination threshold, then it is determined that... for The false nearest neighbor is determined. The default threshold for determining the false nearest neighbor is set to 10, based on the standard threshold recommended by Rosenstein et al. in their original study of the false nearest neighbor method.

[0029] Feature space module statistics in trial values The proportion of pseudo-nearest neighbors among all phase points, as the proportion of pseudo-nearest neighbors increases with the embedding dimension. When the increase first falls below a preset proximity ratio threshold, the feature space module will adjust the current embedding dimension. As the first The embedding dimension parameter of the 3D vibration signal is denoted as . The preset proximity ratio threshold is set to 0.05, based on the principle that the reconstructed phase space can retain a sufficiently rich dynamic structure while eliminating spurious neighbor points.

[0030] Feature space module according to the first Time delay parameters of vibration signal and embedded dimension parameters , for the 3D vibration signal Perform phase space reconstruction to obtain the first... The reconstructed trajectory matrix corresponding to dimension Reconstructing the trajectory matrix Depend on Composed of a number of phase point vectors, , of which The vectors of the phase points are: , This indicates the transpose operation.

[0031] Feature space module for the first The reconstructed trajectory matrix corresponding to dimension Calculate the Euclidean distance for each pair of adjacent trajectory points, where adjacent trajectory points are indexed by their phase points. and The corresponding two phase vectors and The Euclidean distance is calculated as follows: ,in, Representing vectors Norm. The feature space module sums the Euclidean distances between all adjacent trajectory points and divides the sum by the total number of pairs of adjacent trajectory points. The results obtained will be used as the first The average trajectory distance in dimension is denoted as .

[0032] Feature space module for all The average trajectory distance was calculated for each dimension. , will this The average trajectory distances are numbered from 1 to 1 according to the dimension. The sequence is used to construct the vibration space feature vector. The feature space module will use the vibration space feature vectors. Mapping to a preset polar coordinate system, the mapping method is as follows: [Map the vector...] Each element The numerical mapping is the polar angle in the polar coordinate system. polar angle The calculation method is as follows ,in, Vibration space characteristic vector The minimum value of all elements in the set. Vibration space characteristic vector Find the maximum value of all elements in the array; number the dimensions. Mapped to polar radius in polar coordinate system Polar diameter The calculation method is as follows The feature space module sets the polar coordinates of all dimensions. The set of values ​​is used as the vibration characteristic space and output to the parameter adaptation module.

[0033] Example 3: In specific implementation, please refer to Figure 4 The parameter adaptation module receives the vibration feature space output by the feature space module. The vibration feature space is represented by... polar coordinate points The set of points constituted, wherein, For the first The corresponding polar angle of the dimension For the first The corresponding polar radius of the dimension , This represents the total number of dimensions of the multidimensional vibration signal.

[0034] The parameter adaptation module will match all parameters in the vibration characteristic space. The polar coordinates of the nth point are converted from polar coordinates to Cartesian coordinates as follows: The Cartesian x-coordinates corresponding to each polar point Cartesian ordinate The parameter adaptation module will convert the obtained... The geometric center of the Cartesian coordinate points is taken as the center point of the vibration characteristic space, and the x-coordinate of the center point of the vibration characteristic space is... The ordinate of the center point is .

[0035] The parameter adaptation module internally stores multiple preset vibration mode categories, and the total number of vibration mode categories is denoted as . Each vibration mode category has pre-defined category center point coordinates. The category center point coordinates of a vibration mode category are determined through pre-collected sample data. The determination process is as follows: multi-dimensional vibration signal samples are collected under the condition that the spindle is in a known vibration mode type. A sample vibration feature space is generated for the multi-dimensional vibration signal samples using the same method as the feature space module. The geometric center of the transformed Cartesian coordinates of all polar coordinate points in the sample vibration feature space is calculated. The average value of the geometric centers of all samples under the same vibration mode type is taken as the category center point coordinates of that vibration mode category. The vibration mode categories include at least unbalanced vibration mode, misalignment vibration mode, bearing loosening vibration mode, and rubbing vibration mode, and their corresponding category center point coordinates are denoted as follows: , , , .

[0036] The parameter adaptation module calculates the center point of the vibration characteristic space. and Category center point of each vibration mode category Spatial distance between them Spatial distances are calculated using the Euclidean distance method: in, The center point of the vibration characteristic space and the first The spatial distance between the center points of each vibration mode category. The x-coordinate represents the center point of the vibration characteristic space. The ordinate represents the center point of the vibration characteristic space. Indicates the first The x-coordinate of the center point of each vibration mode category. Indicates the first The ordinate of the center point of each vibration mode category. The index number for the vibration mode category. , This represents the total number of vibration mode categories.

[0037] The parameter adaptation module will calculate all The spatial distances are sorted in ascending order to obtain a spatial distance sequence. The parameter adaptation module selects the first, second, and third spatial distances from this sequence, and uses the three vibration mode categories corresponding to these three spatial distances as candidate vibration mode categories. The index numbers of the three candidate vibration mode categories are denoted as follows: , , The corresponding spatial distances are denoted as follows: , , The rationale for selecting three candidate vibration mode categories from three spatial distances is that the smaller the spatial distance, the higher the similarity between the current vibration state and the vibration mode category. Selecting the first three can balance coverage of similar vibration modes and suppression of noise interference.

[0038] The parameter adaptation module internally stores a control strategy library, which contains control strategy nodes corresponding one-to-one with each vibration mode category. Each control strategy node stores a set of stiffness adjustment parameters and a set of damping adjustment parameters. The stiffness and damping adjustment parameters stored in the control strategy nodes are determined through prior simulation and experimentation. The determination method is as follows: for each vibration mode category, the stiffness and damping adjustment parameters are used as optimization variables in the spindle dynamics simulation model. With minimizing the vibration amplitude as the optimization objective, a set of optimal stiffness and damping adjustment parameters is obtained by iteratively solving using a genetic algorithm. The obtained stiffness and damping adjustment parameters are then fine-tuned on a test bench and used as the parameters stored in the control strategy nodes.

[0039] The parameter adaptation module uses the index numbers of the three candidate vibration mode categories. , , Obtain the corresponding control strategy nodes respectively, starting from the first candidate vibration mode category. Extract stiffness adjustment parameters from control strategy nodes and damping adjustment parameters From the second candidate vibration mode category Extract stiffness adjustment parameters from control strategy nodes and damping adjustment parameters From the third candidate vibration mode category Extract stiffness adjustment parameters from control strategy nodes and damping adjustment parameters .

[0040] The parameter adaptation module performs a weighted fusion of all extracted stiffness and damping adjustment parameters based on the reciprocal of their corresponding spatial distances. The weighting method is as follows: the reciprocal of the spatial distance corresponding to the first candidate vibration mode category is... The reciprocal of the spatial distance corresponding to the second candidate vibration mode category is The reciprocal of the spatial distance corresponding to the third candidate vibration mode category is Let the sum of the reciprocals of the three spatial distances be denoted as The weight corresponding to the first candidate vibration mode category is: The weight corresponding to the second candidate vibration mode category is The weight corresponding to the third candidate vibration mode category is .

[0041] The parameter adaptation module weights and fuses the stiffness adjustment parameters according to their weights to obtain the stiffness adjustment parameters in the initial control parameter combination. , The calculation method is as follows The parameter adaptation module weights and fuses the damping adjustment parameters with equal weights to obtain the damping adjustment parameters in the initial control parameter combination. , The calculation method is as follows The parameter adaptation module will and The initial control parameter combination is output to the feedforward correction module.

[0042] Example 4: In specific implementation, please refer to Figure 5 The feedforward correction module receives the initial control parameter combination from the parameter adaptation module and the multi-dimensional vibration signal from the vibration acquisition and judgment module. The feedforward correction module synthesizes the signals of all dimensions in the multi-dimensional vibration signal at each sampling time. The synthesis method involves calculating the root mean square (RMS) value of the sampled vibration signal values ​​for each dimension, and using the RMS value sequence as the instantaneous amplitude sequence reflecting the overall vibration intensity of the principal shaft. The value corresponding to each sampling time in the instantaneous amplitude sequence is denoted as... ,in, For sampling time index, , This represents the total number of sampling points recorded up to and including the current time.

[0043] The feedforward correction module extracts the current time from the instantaneous amplitude sequence. Previous consecutive The instantaneous amplitude values ​​at each historical moment are used to form a sequence of instantaneous amplitude values ​​for fitting, denoted as . . The value is set to 50. The basis for this setting is that at a sampling rate of 1kHz, 50 sampling points correspond to a time window of 50 milliseconds. This time window can contain at least one complete vibration cycle to reflect the instantaneous change trend, while avoiding the trend becoming blurred due to excessive time.

[0044] The feedforward correction module extracts the continuous A polynomial fitting was performed on the instantaneous amplitude sequence segments at each historical moment. The fitting method was least-squares polynomial fitting, and the order of the fitting polynomial was set to 3. This was chosen because a 3rd-order polynomial is sufficient to describe the acceleration variation characteristics of the vibration signal, and non-zero coefficients of higher-order terms can effectively express the local curvature direction of the amplitude variation curve. The resulting instantaneous amplitude variation curve is represented as follows: in, Represents a relative time variable, referring to the time preceding the current moment. A historical moment as ; Indicates relative time The instantaneous amplitude of the fitted value; The constant term coefficients of the fitting polynomial; The coefficients of the first-order terms of the fitted polynomial; The coefficients of the quadratic term in the fitted polynomial; These are the coefficients of the cubic term in the fitted polynomial. Four coefficients. , , , The objective function is obtained by minimizing the sum of squared residuals and solving it using the normal equation method.

[0045] The feedforward correction module uses the fitted instantaneous amplitude change curve as a reference. relative time Find the first and second derivatives, the first derivative... Second derivative The feedforward correction module will use the relative time value corresponding to the current moment. Substituting the values ​​into the expressions for the first and second derivatives, we can calculate the value of the first derivative at the current time. and second derivative value .

[0046] The feedforward correction module is based on the first derivative value and second derivative value Determine the direction of change in the spindle's vibration acceleration. The rule for determining the direction of change in vibration acceleration is: when... and When this occurs, it indicates that not only is the instantaneous amplitude increasing, but the rate of increase is also rising, thus determining the direction of change in vibration acceleration as the direction of acceleration divergence; when and When this occurs, it indicates that not only is the instantaneous amplitude decreasing, but the rate of decrease is also decreasing, thus determining that the direction of change in vibration acceleration is the direction of deceleration convergence; if and If it does not belong to either of the two symbol combinations mentioned above, then the direction of the vibration acceleration change is determined to be maintaining the current state.

[0047] When the direction of vibration acceleration change is the direction of accelerating divergence, the feedforward correction module increases the stiffness and damping adjustment parameters in the initial control parameter combination according to the preset first step length. The preset first step length is set as the stiffness adjustment parameter in the initial control parameter combination. 15% and damping adjustment parameters The 15% setting is based on step response testing of the spindle actuator, which determined that with a 15% parameter increment, the actuator's output reverse damping force can be adjusted within a single control cycle without causing overshoot. The feedforward correction module sets the adjusted stiffness adjustment parameter to... Set the adjusted damping parameters to And will be by and The control commands formed are used as feedforward control commands.

[0048] When the direction of vibration acceleration change is the direction of deceleration convergence, the feedforward correction module decreases the stiffness adjustment parameter and damping adjustment parameter in the initial control parameter combination according to a preset second step size. The preset second step size is set to the stiffness adjustment parameter in the initial control parameter combination. 8% and damping adjustment parameters The 8% setting is based on the principle that during the deceleration convergence phase, to avoid excessively weakening the suppressive force and causing vibration rebound, the parameter reduction should be less than the increase during the acceleration divergence phase. Comparative experiments determined that 8% is a suitable reduction to maintain a stability margin during the convergence phase. The feedforward correction module sets the adjusted stiffness adjustment parameter to... Set the adjusted damping parameters to And will be by and The control commands formed are used as feedforward control commands.

[0049] When the direction of vibration acceleration change is to maintain the current state, the feedforward correction module does not adjust the initial control parameter combination, but directly adjusts the stiffness adjustment parameter in the initial control parameter combination. and damping adjustment parameters The stiffness and damping adjustment parameters in the feedforward control command are used to generate the feedforward control command. The feedforward correction module then sends the final generated feedforward control command to the execution closed-loop module.

[0050] Example 5: In specific implementation, please refer to Figure 6 The closed-loop execution module receives feedforward control commands generated by the feedforward correction module. These feedforward control commands include stiffness adjustment parameters. and damping adjustment parameters The closed-loop execution module converts the feedforward control commands into drive signals, which in turn drive the active actuator on the spindle to output a reverse damping force. The active actuator is an electromagnetic actuator mounted on the spindle bearing housing, and its parameters are adjusted according to stiffness. Adjust the elastic component of the output force according to the damping adjustment parameters. Adjust the damping component of the output force. The elastic component is proportional to the radial displacement of the spindle, and the damping component is proportional to the radial velocity of the spindle.

[0051] While the active actuator outputs a reverse damping force, the execution closed-loop module simultaneously acquires the residual vibration signal after the active actuator outputs the reverse damping force through vibration sensors arranged on the spindle. The residual vibration signal is a multi-dimensional signal, with dimensions consistent with those of the multi-dimensional vibration signal. The execution closed-loop module performs a fast Fourier transform on each dimension of the residual vibration signal to obtain the spectrum of each dimension of the residual vibration signal.

[0052] The closed-loop execution module internally stores a list of preset frequency points, which includes... There are several preset frequency points, selected to cover the spindle's rotation frequency and multiple harmonics thereof, specifically the 1st, 2nd, 3rd, 4th, and 5th harmonics of the spindle's rotation frequency. These preset frequency points are arranged in ascending order of frequency. The closed-loop module extracts the amplitude corresponding to each preset frequency point from the spectrum of the residual vibration signal in each dimension, and then performs root mean square synthesis of the amplitudes of all dimensions at the same preset frequency point to obtain the residual amplitude at that preset frequency point. The residual amplitude values ​​at each preset frequency point are constructed into a residual amplitude vector in ascending order of frequency. This residual amplitude vector is denoted as... ,in Indicates the first The residual amplitude at a preset frequency point , This represents the total number of preset frequency points.

[0053] The closed-loop execution module internally stores a preset residual amplitude tolerance vector, denoted as . ,in, Indicates the first The maximum allowable residual amplitude at a preset frequency point. The residual amplitude tolerance vector is preset based on the basic vibration amplitude spectrum of the spindle at the current speed. The basic vibration amplitude spectrum is obtained as follows: under the condition that the spindle starts each time and no vibration suppression measures are applied, the vibration signal of the spindle is collected during the entire process of acceleration from rest to the target speed. The vibration signal is subjected to short-time Fourier transform, and the spectrum segment after the spindle reaches the steady-state speed is extracted as the basic vibration amplitude spectrum. Extraction is performed from the basic vibration amplitude spectrum. The residual amplitude tolerance value at each preset frequency point is obtained by multiplying the base amplitude at each preset frequency point by a tolerance factor. The tolerance factor is set to 0.2, based on the ratio of the good operating condition level to the upper limit of allowable vibration in the ISO 10816-3 standard for evaluating machine tool spindle vibration. That is, a good suppression effect is considered to be achieved when the vibration amplitude is reduced to 20% of the initial state. The basic vibration amplitude spectrum is updated every time the spindle is started to reflect the baseline vibration characteristics of the spindle under the current operating environment.

[0054] The closed-loop module executes the residual magnitude vector With residual magnitude tolerance vector Perform element-by-element comparison, the comparison method is from index Start sequentially to ,Compare and The size relationship. If a certain index exists. satisfy Then the closed-loop module is executed to determine the index. The corresponding preset frequency point is the frequency point that exceeds the set frequency.

[0055] The closed-loop module is executed to obtain the excess amplitude difference corresponding to all frequency exceedance points. For each frequency exceedance point, the excess amplitude difference is calculated as follows: The closed-loop module sums the amplitude differences at all frequency points exceeding the limit, and uses this sum as the residual vibration excess value. The formula for calculating the residual vibration excess value is as follows: in, This indicates the excess value of residual vibration. This represents the set of index numbers of all points exceeding the frequency range. Represents the th element in the residual magnitude vector The residual amplitude at a preset frequency point The residual magnitude tolerance vector represents the first... Residual amplitude tolerance at a preset frequency point This is the index number for the preset frequency point.

[0056] The closed-loop module executes based on the residual vibration excess value. Generate correction factors. The correction factors are generated by taking the excess residual vibration value... Compared with the preset excess reference value Comparison, Set to 10% of the sum of the residual amplitude tolerance values ​​at all preset frequency points. If Then the correction factor will be... Set as ;like Then the correction factor will be... Set as The square root value. The setting is based on the following: when the residual vibration exceeds 10% of the total tolerance value, it indicates that the residual vibration exceeds the tolerance range at multiple frequency points, and the correction force needs to be increased; when the residual vibration exceeds 10% of the total tolerance value, linear correction is used to maintain stable control.

[0057] The closed-loop module will correct the coefficients. stiffness adjustment parameters in feedforward control commands Multiplying them yields the stiffness adjustment parameter in the closed-loop control signal. ; Adjust the coefficient Damping adjustment parameters in feedforward control commands Multiplying them yields the damping adjustment parameter in the closed-loop control signal. The closed-loop module will include stiffness adjustment parameters. and damping adjustment parameters The closed-loop control signal is output to the active actuator, and the active actuator adjusts the output reverse resistance force according to the closed-loop control signal.

[0058] 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. A spindle vibration active suppression and control system, characterized in that, The system includes: The vibration acquisition and judgment module acquires multi-dimensional vibration signals of the spindle in operation and performs feature separation operations in the time and frequency domains. Based on the separation results, it determines whether the current vibration state of the spindle meets the preset active suppression intervention conditions. The feature space module, when the active suppression intervention condition is met, performs nonlinear dynamic feature extraction on the multidimensional vibration signal, and constructs the vibration feature space of the principal shaft based on the distribution law of the extracted nonlinear dynamic features in the reconstructed phase space. The parameter adaptation module dynamically matches the initial control parameter combination corresponding to the current vibration state from the control strategy library based on the projection intensity of the vibration feature space on multiple preset vibration mode categories. The feedforward correction module dynamically corrects the initial control parameter combination based on the instantaneous change trend of the multi-dimensional vibration signal, and generates feedforward control commands. The closed-loop module executes the feedforward control command to drive the active actuator on the spindle to output a reverse damping force, and simultaneously collects the residual vibration signal after the active actuator outputs the reverse damping force. Based on the residual vibration signal, the feedforward control command is corrected in a closed loop to generate a closed-loop control signal.

2. The active suppression and control system for spindle vibration according to claim 1, characterized in that, The vibration sampling and judgment module performs feature separation operations in the time domain and frequency domain, including: Empirical mode decomposition is performed on the multidimensional vibration signal to obtain multiple intrinsic mode components. The multiple intrinsic mode components are then sorted in descending order of frequency to obtain a set of high-frequency components and a set of low-frequency components. Calculate the instantaneous energy value of each intrinsic mode component in the high-frequency component set, and take the average of all instantaneous energy values ​​as the high-frequency instantaneous energy average; calculate the instantaneous energy value of each intrinsic mode component in the low-frequency component set, and take the average of all instantaneous energy values ​​as the low-frequency instantaneous energy average. The ratio of the average high-frequency instantaneous energy to the average low-frequency instantaneous energy is used as the high-frequency proportion coefficient of vibration energy, and it is determined whether the high-frequency proportion coefficient of vibration energy is greater than a preset high-frequency energy proportion threshold. If the high-frequency proportion coefficient of the vibration energy is greater than the high-frequency energy proportion threshold, then the spindle is determined to have reached the active suppression intervention condition.

3. The active suppression and control system for spindle vibration according to claim 1, characterized in that, The feature space module performs nonlinear dynamic feature extraction on the multi-dimensional vibration signal, including: For each dimension of the multi-dimensional vibration signal, the time delay parameter of each dimension of the vibration signal is determined by the mutual information method, and the embedding dimension parameter of each dimension of the vibration signal is determined by the pseudo-nearest neighbor method. Based on the time delay parameter and the embedding dimension parameter of the vibration signal in each dimension, the phase space of the vibration signal in each dimension is reconstructed to obtain the reconstructed trajectory matrix corresponding to each dimension. Calculate the Euclidean distance between all adjacent trajectory points in the reconstructed trajectory matrix corresponding to each dimension, and take the mean of the Euclidean distances of all adjacent trajectory points as the average trajectory distance for each dimension; The average trajectory distances of all dimensions are constructed into vibration space feature vectors according to the dimension numbering order, and the vibration space feature vectors are mapped to a preset polar coordinate system to obtain the vibration feature space.

4. The active suppression and control system for spindle vibration according to claim 1, characterized in that, The parameter adaptation module dynamically matches an initial control parameter combination corresponding to the current vibration state from a control strategy library based on the projection intensity of the vibration feature space across multiple preset vibration mode categories, including: Calculate the spatial distance between the center point of the vibration feature space and the center points of each of the preset multiple vibration mode categories, and sort all the calculated spatial distances in ascending order to obtain a spatial distance sequence; The three vibration mode categories corresponding to the top three spatial distances in the spatial distance sequence are selected as candidate vibration mode categories. For each of the candidate vibration mode categories, obtain the corresponding control strategy node in the control strategy library. Extract the stiffness adjustment parameter and damping adjustment parameter from each control strategy node. Then, weight and fuse all the extracted stiffness adjustment parameters and damping adjustment parameters according to the reciprocal of the corresponding spatial distance to obtain the initial control parameter combination.

5. The active suppression and control system for spindle vibration according to claim 1, characterized in that, The feedforward correction module dynamically corrects the initial control parameter combination based on the instantaneous change trend of the multi-dimensional vibration signal, including: The instantaneous amplitude sequence of the multidimensional vibration signal at multiple consecutive historical moments before the current moment is obtained, and the instantaneous amplitude sequence is fitted with a polynomial to obtain the instantaneous amplitude change curve. Calculate the first and second derivatives of the instantaneous amplitude change curve at the current moment, and determine the direction of vibration acceleration change of the main shaft based on the first and second derivatives. If the direction of the vibration acceleration change is an accelerating divergent direction, then the stiffness adjustment parameter and damping adjustment parameter in the initial control parameter combination are increased according to the preset first step length to obtain the feedforward control command; If the direction of the change in vibration acceleration is the direction of deceleration convergence, then the stiffness adjustment parameter and damping adjustment parameter in the initial control parameter combination are reduced according to the preset second step size to obtain the feedforward control command.

6. The active suppression and control system for spindle vibration according to claim 1, characterized in that, The closed-loop execution module performs closed-loop correction of the feedforward control command based on the residual vibration signal, including: The residual amplitude of the residual vibration signal at multiple preset frequency points is obtained, and the residual amplitude at the multiple preset frequency points is constructed into a residual amplitude vector in order of frequency from low to high. The residual amplitude vector is compared element by element with the preset residual amplitude tolerance vector to determine whether there is an out-of-frequency point in the residual amplitude vector that is greater than the corresponding element in the residual amplitude tolerance vector. If the frequency exceeds the specified frequency, then obtain the difference in amplitude exceeding all the frequency exceeds the specified frequency, and use the sum of all the difference in amplitude exceeding the specified frequency as the residual vibration excess value. A correction coefficient is generated based on the residual vibration excess value, and the correction coefficient is multiplied by the stiffness adjustment parameter and damping adjustment parameter in the feedforward control command to obtain the closed-loop control signal.

7. The active suppression and control system for spindle vibration according to claim 3, characterized in that, When using the mutual information method to determine the time delay parameter of the vibration signal in each dimension, the time delay corresponding to the first local minimum value of the mutual information function is selected as the time delay parameter.

8. The active suppression and control system for spindle vibration according to claim 3, characterized in that, When using the pseudo-nearest neighbor method to determine the embedding dimension parameter of the vibration signal in each dimension, when the proportion of pseudo-nearest neighbor points first falls below a preset proximity ratio threshold as the embedding dimension increases, the current embedding dimension is used as the embedding dimension parameter.

9. The active suppression and control system for spindle vibration according to claim 4, characterized in that, In the weighted fusion, the weight corresponding to each stiffness adjustment parameter and the damping adjustment parameter is the ratio of the reciprocal of the corresponding spatial distance to the sum of the reciprocals of all spatial distances.

10. The active suppression and control system for spindle vibration according to claim 6, characterized in that, The residual amplitude tolerance vector is preset based on the basic vibration amplitude spectrum of the spindle at the current speed, and the basic vibration amplitude spectrum is updated each time the spindle is started.

Images