A real-time monitoring and automatic suppression method for spindle vibration of a numerical control machine tool

Abstract

The application provides a numerical control machine tool spindle vibration real-time monitoring and automatic inhibition method, comprising: acquiring the spindle vibration parameters of a titanium alloy artificial hip joint femoral stem bionic structure in the machining process of a numerical control machine tool, the vibration parameters including vibration frequency and vibration amplitude, and simultaneously collecting the porosity distribution information of a machining area; according to the porosity distribution information, the machining area is divided into a surface loose bone structure area and a deep dense bone structure area, the vibration parameters in the machining process of different areas are extracted respectively, and a mapping relationship between the vibration parameters and the porosity distribution is established; according to the vibration parameter change rule of the deep dense bone structure area, a cooperative change model of the vibration parameters and the cutting state is established, and the vibration trend of the spindle when machining in a low porosity area is predicted.

Description

Technical Field

[0001] This invention relates to the field of information technology, and in particular to a method for real-time monitoring and automatic suppression of spindle vibration in CNC machine tools. Background Technology

[0002] When machining the biomimetic femoral stem structure of a titanium alloy artificial hip joint, the porosity of the material exhibits a gradient change from the surface porous bone structure to the deep dense bone structure. This difference in material structure leads to varying vibration responses of the spindle during machining in different regions. As the spindle cuts from the high-porosity surface porous bone structure and gradually penetrates into the low-porosity dense bone structure, the material removal rate, cutting force, and tool-workpiece contact state all dynamically change. This dynamic change in cutting conditions causes drastic fluctuations in the spindle vibration frequency and amplitude, especially during continuous machining in the low-porosity dense bone structure region, where the coupling effect between vibration response and cutting state is more significant. The nonlinear evolution of spindle vibration leads to instability in the machining process, severely affecting the machining accuracy and surface quality of the biomimetic structure. Therefore, it is urgent to study the laws governing the synergistic changes in spindle vibration and cutting state, and to develop an automatic spindle vibration suppression technology based on real-time detection and feedback control of the machining state to achieve high-quality machining of the titanium alloy artificial hip joint femoral stem biomimetic structure. Summary of the Invention

[0003] This invention provides a method for real-time monitoring and automatic suppression of spindle vibration in CNC machine tools, mainly including:

[0004] The spindle vibration parameters of the biomimetic structure of the femoral stem of the titanium alloy artificial hip joint were obtained during the CNC machine tool machining process. The vibration parameters included vibration frequency and vibration amplitude. At the same time, the porosity distribution information of the machining area was collected.

[0005] Based on the porosity distribution information, the processing area is divided into a surface loose bone structure area and a deep dense bone structure area. Vibration parameters during the processing of different areas are extracted, and a mapping relationship between vibration parameters and porosity distribution is established.

[0006] Analyze the vibration parameter variation trend in the surface porous bone structure region, and determine whether the vibration parameter exhibits nonlinear fluctuation with the change of porosity based on the vibration parameter variation trend. If so, and the vibration amplitude exceeds the preset safety threshold, the preset vibration suppression mechanism is immediately triggered.

[0007] Vibration parameters of deep dense bone structure regions during continuous machining are extracted to determine whether the vibration parameters exhibit coordinated changes with the cutting state. If the vibration frequency is positively correlated with the cutting force, the variation law of vibration parameters under the cutting state is recorded.

[0008] Based on the variation law of vibration parameters in the deep dense bone structure region, a collaborative variation model of vibration parameters and cutting state is established to predict the vibration trend of the spindle when machining in the low porosity region.

[0009] Vibration parameters are feature extracted to extract key features such as vibration amplitude, vibration frequency, and the rate of change of vibration parameters over time. Combined with the cooperative change model, it is determined whether the vibration amplitude continues to increase and the vibration frequency tends to stabilize. If so, a preset automatic suppression mechanism is triggered.

[0010] The automatic suppression mechanism adjusts the spindle speed and feed rate to reduce vibration amplitude and stabilize vibration frequency. At the same time, it monitors the changes in cutting status in real time, re-collects vibration parameters, and judges whether the vibration suppression effect meets the preset requirements. If the vibration amplitude drops below the threshold and the vibration frequency tends to be stable, the automatic suppression process is completed.

[0011] Furthermore, the acquisition of spindle vibration parameters during the CNC machine tool machining process of the biomimetic femoral stem structure of the titanium alloy artificial hip joint includes vibration frequency and vibration amplitude. Simultaneously, porosity distribution information of the machining area is collected, including:

[0012] A spindle speed sensor is used to acquire spindle rotation frequency data in the machining area of ​​the femoral stem of the titanium alloy artificial hip joint. A vibration sensor collects the spindle vibration frequency and amplitude. Strain gauges are placed at the machining point to measure spindle torque, and spindle power parameters are read from the CNC system. A parameter correlation matrix is ​​established based on the acquired spindle rotation frequency, vibration frequency, and vibration amplitude data to evaluate spindle speed stability, thereby creating a vibration parameter distribution map in the machining area. An optical sensor monitors the surface roughness of the cutting point in real time. An acoustic emission sensor collects acoustic emission signals during machining, and a mapping relationship is established between the acoustic emission signal characteristic values ​​and the tool feed rate and depth of cut to monitor the machined surface quality online. The porosity distribution of the machining area is obtained by combining the spindle vibration parameter distribution map. A preset threshold is established based on the titanium alloy material properties. The monitoring results are compared with the preset threshold to determine the machining status. If the acoustic emission signal characteristic value exceeds the preset range, the tool feed rate and depth of cut are adjusted until the acoustic emission signal characteristic value returns to the preset range.

[0013] Furthermore, based on the porosity distribution information, the processing area is divided into a surface porous bone structure region and a deep dense bone structure region. Vibration parameters during the processing of different regions are extracted, and a mapping relationship between the vibration parameters and the porosity distribution is established, including:

[0014] Point cloud datasets were collected from the processing area. Density clustering algorithms were applied to the point cloud data. Based on the porosity distribution density, point sets of loosely structured and densely structured regions were marked in the point cloud dataset. Boundary calibration was performed on the point set distribution based on region boundary determination rules. The width of the structural transition zone between the surface loose bone structure region and the deep dense bone structure region was calculated based on the boundary calibration results. For the point sets within the surface region, vibration frequency and amplitude parameters were collected using vibration sensors. The vibration parameters were normalized to obtain the surface vibration feature vector. The depth detector measures the bone structure depth parameters in the point cloud dataset of the measurement area. Support vector regression is used to establish a mapping relationship between vibration feature vectors and porosity. Vibration frequency parameters and vibration amplitude parameters are collected from the deep region. The vibration parameters are normalized to obtain the deep vibration feature vector. A mapping function is constructed based on the surface vibration feature vector and the deep vibration feature vector. A porosity threshold is extracted from the mapping function. When the porosity of the detection point is higher than the threshold, it is classified into the surface loose bone structure region. When the porosity of the detection point is lower than the threshold, it is classified into the deep dense bone structure region.

[0015] Furthermore, the analysis of the vibration parameter variation trend in the surface porous bone structure region determines whether the vibration parameters exhibit nonlinear fluctuations with changes in porosity. If so, and the vibration amplitude exceeds a preset safety threshold, a preset vibration suppression mechanism is immediately triggered, including:

[0016] Vibration frequency and amplitude data were collected in the surface porous bone structure area using vibration sensors. Fourier transform was used to perform time-frequency analysis on the vibration frequency data to generate a vibration frequency spectrum, from which the dominant frequency component and harmonic components were extracted. The vibration frequency change rate curve was calculated based on the dominant frequency and harmonic components. A Kalman filter was used to filter the vibration amplitude data to obtain the vibration amplitude curve, with the fluctuation detection period set as continuous sampling data. The maximum and minimum rates of change within the fluctuation period were extracted from the vibration frequency change rate curve, and the fluctuation range was calculated. A preset nonlinearity threshold was used to determine whether the vibration frequency exhibited nonlinear fluctuations. Porosity data was collected using vibration sensors, and a recurrent neural network was used to establish a mapping relationship between the vibration frequency change rate and porosity changes. The recurrent neural network's input layer is a sequence of vibration frequency change rates, its hidden layers are three fully connected layers, and its output layer is the porosity change value. The maximum and minimum amplitudes within the fluctuation detection period are extracted from the vibration amplitude curve, and the vibration amplitude fluctuation range is calculated. When the fluctuation range exceeds a preset safety threshold, a vibration suppression mechanism is triggered. A suppression parameter mapping table is established based on the vibration frequency change rate and the vibration amplitude fluctuation range. The suppression parameters include spindle speed adjustment and feed rate adjustment. When the vibration suppression mechanism is triggered, the corresponding adjustment parameters are retrieved from the mapping table. A proportional-integral controller is used to dynamically adjust the spindle speed and feed rate. The controller input is the deviation between the vibration amplitude and the safety threshold, and the output is the actual speed and feed rate adjustment, until the vibration amplitude drops below the safety threshold.

[0017] Furthermore, the vibration parameters of the deep, dense bone structure region are extracted during continuous machining, and it is determined whether the vibration parameters exhibit coordinated changes with the cutting state. If the vibration frequency is positively correlated with the cutting force, the variation law of the vibration parameters under this cutting state is recorded, including:

[0018] Vibration and cutting force data were collected in deep, dense bone structures using vibration and cutting force sensors. Wavelet transform was used to decompose the vibration frequency data, obtaining the frequency and amplitude components. The mean and standard deviation of the cutting force were calculated based on the cutting force data. When the standard deviation was less than 10% of the mean, a stable cutting state was defined, and the vibration frequency and amplitude components within this range were recorded. The correlation between vibration frequency and cutting force was calculated using the Pearson correlation coefficient, and the vibration parameter fluctuations of each dense layer of the deep bone structure were extracted from the vibration frequency variation trend. Depth data for each layer of the deep bone structure was acquired using a depth sensor; when the depth value exceeded a preset depth threshold, it was determined to be deep, dense bone. In dense structural regions, a bone structure density distribution curve is constructed based on depth data. Tool wear data is acquired from cutting tool wear sensors, and a long short-term memory network is used to establish a temporal mapping relationship between cutting force and vibration frequency. The network input is a cutting force sequence and a vibration frequency sequence, and the output is a co-variation feature value. When the co-variation feature value is greater than zero and the ratio of the vibration frequency increment to the cutting force increment is greater than a preset threshold, it is determined to be a positive correlation state, and the vibration frequency, vibration amplitude, cutting force data, and tool wear data in this state are recorded. Based on the recorded data, a cutting state parameter mapping table is constructed, which includes four parameter dimensions: vibration frequency, vibration amplitude, cutting force, and tool wear, and the correspondence between the parameters is established.

[0019] Furthermore, based on the variation law of vibration parameters in the deep dense bone structure region, a coordinated variation model of vibration parameters and cutting state is established to predict the vibration trend of the spindle during machining in a low porosity region, including:

[0020] Vibration sensors were used to collect raw vibration parameter data in deep, dense bone structures. Cutting state parameters were obtained through a cutting state sensor. Porosity values ​​in low-porosity regions were read from an ultrasonic porosity analyzer. The raw vibration parameter data were normalized using a range normalization method. A short-time Fourier transform was calculated based on the normalized vibration parameter data to obtain the time-frequency feature matrix of the vibration parameters. Vibration frequency and amplitude features were extracted from the time-frequency feature matrix. Spindle speed data was acquired using a spindle speed sensor, and cutting force data collected by a cutting force sensor was recorded. A multi-dimensional feature vector containing vibration features, speed data, and cutting force data was constructed. A recurrent neural network was then used to build... A collaborative variation model of vibration parameters and cutting state is established. The input layer of the model is a multidimensional feature vector, the hidden layer uses long short-term memory units, and the output layer is the predicted value of vibration parameter change. A Bayesian regressor is trained using the output value of the collaborative variation model and the measured porosity value. The input of the regressor is the predicted value of vibration parameter change, and the output is the vibration trend prediction curve. When the ratio of the vibration frequency increment to the cutting force increment in the prediction curve is greater than a preset threshold, it is determined that the vibration parameters and the cutting state are positively correlated, and the predicted value of vibration parameters under this state is recorded. The vibration trend prediction curve is fitted using cubic spline interpolation, and the slope of the fitted curve in the low porosity region is calculated to obtain the variation trend of the spindle vibration parameters.

[0021] Furthermore, the vibration parameters are feature extracted to obtain key features such as vibration amplitude, vibration frequency, and the rate of change of vibration parameters over time. Combined with a collaborative change model, it is determined whether the vibration amplitude continues to increase and the vibration frequency tends to stabilize. If so, a preset automatic suppression mechanism is triggered, including:

[0022] Vibration data sequences are continuously acquired using vibration sensors. Wavelet transform is used to perform multi-scale decomposition of the vibration data sequences, extracting vibration amplitude, frequency, and parameter time-varying rate of change from the decomposition results. The frequency variance within a time window is calculated based on the vibration frequency data. A Kalman filter is used to filter the vibration amplitude data, and the amplitude change rate within a continuous sampling period is calculated from the filtered amplitude. A vibration parameter predictor is constructed using a recurrent neural network. The predictor's input includes vibration frequency, amplitude, and parameter time-varying rate of change, and its output is the trend of vibration parameter changes over time. A collaborative change discriminant is used to calculate the vibration frequency... The vibration frequency variance and the rate of change of vibration amplitude exhibit a coordinated change characteristic. When the vibration frequency variance is less than the stability threshold and the rate of change of vibration amplitude is greater than the growth threshold, the suppression controller is triggered. The suppression controller selects the corresponding spindle speed adjustment and feed rate adjustment based on the coordinated change characteristic, forming a suppression control command. The control command is executed to adjust the machining parameters, and a proportional-integral controller is used to achieve closed-loop control of the speed and feed rate. The output gain of the controller is calculated based on the rate of change of vibration amplitude. Real-time vibration data during the suppression process is collected from the vibration sensor. When the rate of change of vibration amplitude drops below the growth threshold, the current suppression control parameters remain unchanged.

[0023] Furthermore, the automatic suppression mechanism adjusts the spindle speed and feed rate to reduce vibration amplitude and stabilize vibration frequency, while simultaneously monitoring changes in the cutting state in real time, re-collecting vibration parameters, and determining whether the vibration suppression effect meets the preset requirements. If the vibration amplitude drops below the threshold and the vibration frequency tends to stabilize, the automatic suppression process is completed, including:

[0024] A suppression parameter selector is used to obtain the speed and feed adjustment values ​​from a preset parameter mapping table. Parameter adjustment commands are sent to the spindle controller and feed controller via a CNC communication interface. The actual speed and feed rate are calculated based on the position feedback signals from the spindle encoder and feed encoder. A proportional-integral controller (PIC) is used to achieve closed-loop control, with the PLC coefficients automatically adjusted according to the vibration amplitude. Vibration frequency and amplitude data are collected during the adjustment process using a vibration sensor. A Kalman filter is used to filter the collected data, and the state noise covariance and measurement noise covariance of the filter are automatically updated according to the sampling period. A recurrent neural network is used to construct a vibration characteristic... The feature extractor's network input layer contains vibration frequency and amplitude sequences, the hidden layer uses long short-term memory units, and the output layer generates vibration frequency stability and vibration amplitude change rate. It acquires cutting state data from a cutting force sensor, establishes a mapping relationship between vibration characteristics and cutting force, and calculates the stability range of vibration parameters based on this mapping relationship. When the vibration amplitude is lower than a preset amplitude threshold and the vibration frequency stability is higher than a preset stability threshold, the current machining parameters remain unchanged, and the vibration and cutting parameters under this state are recorded. If the vibration parameters exceed the stability range, new speed and feed adjustments are calculated based on the vibration characteristics, and the parameter adjustment process is repeated until the vibration parameters enter the stability range.

[0025] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects:

[0026] This invention discloses a method for real-time monitoring and automatic suppression of CNC machine tool spindle vibration. The method acquires spindle vibration parameters and porosity distribution information during machining, divides the machining area into a surface porous bone structure region and a deep dense bone structure region, and establishes a mapping relationship between vibration parameters and porosity distribution. This invention analyzes the variation trend of vibration parameters in different regions, determines whether nonlinear fluctuations or coordinated change characteristics occur, and establishes a coordinated change model between vibration parameters and cutting conditions. When abnormal vibration is detected, this invention triggers a preset automatic suppression mechanism, adjusting the spindle speed and feed rate to reduce the vibration amplitude and stabilize the vibration frequency. This method can effectively monitor and suppress abnormal vibration during machining, improving the machining accuracy and quality of titanium alloy artificial hip joint femoral stems. Attached Figure Description

[0027] Figure 1 This is a flowchart of a method for real-time monitoring and automatic suppression of spindle vibration in CNC machine tools according to the present invention.

[0028] Figure 2 This is a schematic diagram of a method for real-time monitoring and automatic suppression of spindle vibration in CNC machine tools according to the present invention. Detailed Implementation

[0029] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] like Figure 1-2 This embodiment of a method for real-time monitoring and automatic suppression of CNC machine tool spindle vibration may specifically include:

[0031] S101. When machining the biomimetic structure of the femoral stem of a titanium alloy artificial hip joint on a CNC machine tool, the feeding end collects the spindle vibration parameters in real time, including vibration frequency and vibration amplitude. At the same time, it acquires the porosity distribution data of the machining area and establishes the correlation between vibration characteristics and machining state based on these data to support subsequent analysis and suppression operations.

[0032] In this embodiment of the invention, the delivery end achieves comprehensive data acquisition through the collaborative operation of multiple sensors. The acquisition of spindle vibration parameters relies on high-precision sensors, while the extraction of porosity distribution data is combined with the analysis of the physical characteristics of the processing area. Specific operating methods can be flexibly adjusted according to the actual processing scenario, without excessive limitations. For example, when processing transitions from a surface porous bone structure to a deep dense bone structure, the spindle vibration characteristics will exhibit different response patterns due to changes in material porosity. The delivery end must ensure the real-time nature and accuracy of the data to provide a reliable basis for subsequent vibration suppression.

[0033] S1011 The feeding end uses a spindle speed sensor to collect spindle rotation frequency data, and at the same time obtains vibration frequency and vibration amplitude through a vibration sensor. Combined with strain gauge measurement of spindle torque data and spindle power parameters read by the CNC system, a multi-dimensional data set is formed for analyzing the spindle operating status.

[0034] In this embodiment of the invention, the spindle speed sensor employs the Hall effect principle, capturing spindle rotation pulse signals at fixed time intervals and converting them into frequency values. For example, when the spindle speed is set to 3000 revolutions per minute, the sensor collects data with a period of 20 milliseconds, calculating a rotational frequency of approximately 50 Hz. The vibration sensor is a piezoelectric accelerometer, with a measurement range covering 0 to 500 Hz and a sensitivity of 100 millivolts per gravitational acceleration. During processing, a vibration frequency of approximately 156 Hz and an amplitude of 0.15 mm were detected. Strain gauges measure torque using a Wheatstone bridge circuit, with a sensitivity of 2 millivolts per Newton-meter and a measured average torque of 120 Newton-meters. These data collectively constitute a multidimensional characterization of the spindle's operating state, laying the foundation for subsequent analysis.

[0035] S1012. The feeding end constructs a parameter correlation matrix based on the collected spindle rotation frequency, vibration frequency and vibration amplitude data, and generates a vibration parameter distribution map based on the matrix. At the same time, it uses an acoustic emission sensor to collect acoustic emission signals during the processing, and establishes a mapping relationship between the signal characteristics and porosity distribution data to determine the processing status.

[0036] In this embodiment of the invention, the parameter correlation matrix is ​​generated by calculating the correlation coefficient between the spindle rotation frequency, vibration frequency, and vibration amplitude. A correlation coefficient close to 1 indicates a high degree of coupling between parameters. The generated vibration parameter distribution map visually reflects the changes in vibration characteristics of the machining area, such as the transition region from a surface porous bone structure to a deep dense bone structure. The acoustic emission sensor uses a piezoelectric ceramic element with a frequency response range of 100 kHz to 1 MHz. The acquired signal amplitude and energy characteristics are closely related to the tool cutting state. For example, when the feed rate is 200 mm / min and the cutting depth is 0.5 mm, the root mean square value of the acoustic emission signal reaches 0.5 volts, and the event count is 800 times per second. By comparing with porosity distribution data, the changes in material properties of the machining area can be further verified.

[0037] S1013. The feeding end presets the acoustic emission signal threshold according to the characteristics of titanium alloy material, and monitors in real time whether the acoustic emission signal exceeds the threshold range. If it exceeds the threshold range, the signal value is restored to the threshold range by adjusting the tool feed speed and cutting depth, thereby initially stabilizing the machining state.

[0038] In this embodiment of the invention, based on the processing characteristics of titanium alloy TC4, the preset threshold for the root mean square (RMS) value of the acoustic emission signal is 0.8 volts. When the detected RMS value exceeds this threshold, for example, reaching 0.9 volts, it indicates an abnormal cutting condition, possibly due to a local increase in porosity or accelerated tool wear. At this time, the feed end automatically reduces the feed rate from 200 mm / min to 150 mm / min and the cutting depth from 0.5 mm to 0.3 mm until the signal value drops below 0.5 volts. After adjustment, the porosity distribution in the processing area tends to be uniform, with the average value maintained at around 3%, initially achieving stability in the processing state. In this embodiment of the invention, through the above steps, the feed end can comprehensively collect spindle vibration parameters and porosity distribution data, and establish a mapping relationship between the two. This method fully utilizes the advantages of multi-sensor collaborative monitoring, providing data support for subsequent vibration trend prediction and automatic suppression. At the same time, the strategy of real-time adjustment of cutting parameters effectively reduces vibration interference during processing, improving the quality and consistency of the processed surface. Subsequent steps can further optimize the suppression mechanism based on this data, which will not be elaborated further.

[0039] S102. Based on the porosity distribution information of the processing area, the feeding end divides the processing area of ​​the biomimetic structure of the femoral stem of the titanium alloy artificial hip joint into a surface porous bone structure area and a deep dense bone structure area. The main shaft vibration parameters of different areas are extracted and their mapping relationship with the porosity distribution is established. At the same time, the region boundary is divided and the vibration characteristics are quantified through cluster analysis and regression modeling, providing a data basis for subsequent vibration suppression.

[0040] In this embodiment of the invention, the delivery end first acquires porosity distribution data of the processing area and uses this data to structurally divide the processing area. The division process relies not only on the porosity itself but also on the variation patterns of vibration parameters to ensure the accuracy of the region division and its practical processing significance. The divided regions will be used to analyze the vibration characteristics of the surface and deep structures, respectively, and the mapping relationship will provide a basis for vibration suppression strategies. This step does not impose excessive restrictions on specific sensor types or algorithm details; the implementation method can be adjusted by technical personnel according to actual needs.

[0041] S1021. The delivery end collects point cloud datasets of the processing area and processes the point cloud data using a density clustering method. Based on the density characteristics of the porosity distribution, it marks the point sets of the surface loose bone structure region and the point sets of the deep dense bone structure region, and calculates the region boundaries based on the gradient method to complete the structural division. In this embodiment of the invention, the delivery end uses a high-resolution scanning device to collect point cloud data. For example, within a 10 mm × 10 mm processing area, 100,000 discrete points are generated with a sampling interval of 0.1 mm. The density clustering algorithm calculates the local density of each point and the distance between adjacent points, classifying points with a porosity greater than 15% as loose structure point sets and points with a porosity less than 5% as dense structure point sets. Boundary determination uses the gradient method, calculating the porosity change rate between adjacent points. When the change rate exceeds 0.1 mm, it is identified as a region boundary point. This method can effectively capture the transition characteristics of porosity from high to low, and the measured width of the transition zone is approximately 2 mm, providing a spatial basis for subsequent parameter extraction.

[0042] S1022. For the surface porous bone structure area, the feeding end uses vibration sensors to collect vibration frequency and vibration amplitude data during the processing. These data are normalized to generate surface vibration feature vectors. At the same time, the depth of the bone structure is measured by a depth detector. A mapping model between the vibration feature vector and the porosity distribution is constructed by the support vector regression method.

[0043] In this embodiment of the invention, the vibration sensor collects vibration frequencies ranging from 150 to 200 Hz and vibration amplitudes between 0.2 and 0.3 mm in the surface region, reflecting the sensitivity of the low elastic modulus (approximately 5 GPa) of the porous structure to cutting force fluctuations. Normalization maps the raw data to the 0-1 range, facilitating subsequent modeling. The depth probe uses ultrasonic technology with a transmission frequency of 20 MHz, measuring relatively small depth variations in the surface structure, typically less than 5 mm. Support vector regression fits the relationship between vibration frequency, amplitude, and porosity using a kernel function (such as a radial basis function). After model training, the correlation coefficient between vibration amplitude and porosity is as high as 0.91. This mapping model can deduce the porosity distribution from vibration data, providing auxiliary verification for region division.

[0044] S1023. The delivery end collects vibration parameters in the deep dense bone structure region, normalizes the vibration frequency and vibration amplitude to generate a deep vibration feature vector, and constructs a mapping function based on the surface and deep vibration feature vectors to extract the porosity threshold for region division judgment.

[0045] In this embodiment of the invention, the vibration frequency distribution in the deep region is between 80 and 120 Hz, and the vibration amplitude is between 0.05 and 0.1 mm, exhibiting strong vibration suppression characteristics. This is closely related to the high compressive strength (200 to 300 MPa) of this region. The normalized deep vibration feature vector and the surface vector are jointly input into a mapping function, which can be constructed using polynomial fitting or neural network methods. The porosity threshold extracted from this function is typically between 5% and 15%. For example, when the porosity of the detection point is higher than 10%, it is classified as a loose surface region; when it is lower than 10%, it is classified as a dense deep region. This threshold extraction method combines the physical meaning of the vibration parameters with the mechanical properties of the processed area, ensuring the reliability of the classification results.

[0046] S1024. The delivery end verifies the mapping relationship between vibration parameters and porosity distribution through periodic loading experiments, optimizes the region division by utilizing the vibration gradient characteristics of the boundary region, and analyzes the differences in vibration response characteristics of different regions to support the effectiveness of the mapping model.

[0047] In this embodiment of the invention, the feeding end applies a periodic load of 50 to 200 Newtons to the processing area, and records the vibration response of the surface and deep regions. It was found that the vibration amplitude in the surface region is 3 to 4 times that in the deep region, verifying the significant influence of porosity on vibration. The vibration frequency gradient in the boundary region is approximately 50 Hz per millimeter, and the vibration amplitude gradient is approximately 0.1 mm per millimeter. These transitional features are used to optimize the boundary determination rules. The effectiveness of the mapping model is further confirmed by experimental data. For example, when a vibration frequency exceeding 160 Hz and an amplitude greater than 0.25 mm are detected in the surface region, it is accurately identified as a loose structure, while when the frequency in the deep region is below 100 Hz and the amplitude is less than 0.08 mm, it is accurately classified as a dense structure. This difference analysis not only enhances the predictive ability of the model but also provides theoretical support for subsequent suppression strategies. In this embodiment of the invention, the feeding end achieves accurate division of the processing area and feature extraction of vibration parameters through the above steps. The establishment of the mapping relationship fully utilizes the advantages of density clustering and support vector regression, making the correlation between porosity distribution and vibration characteristics clearer. This method can adapt to dynamic changes in the processing of titanium alloy materials, providing a solid data foundation for real-time monitoring and suppression. The application scenarios of the model can be further expanded according to actual needs.

[0048] S103. The feeding end analyzes the changing trend of the spindle vibration parameters in the surface porous bone structure area. Through time-frequency analysis and filtering, it determines whether the vibration parameters exhibit nonlinear fluctuations due to changes in porosity. When the vibration amplitude exceeds the preset safety threshold, it immediately triggers the vibration suppression mechanism to stabilize the processing.

[0049] In this embodiment of the invention, the delivery end acquires and processes vibration data from the surface area in real time to identify its dynamic variation patterns. The analysis process combines characteristics of the frequency and amplitude domains to ensure high accuracy in judging nonlinear vibrations. If abnormal fluctuations are detected, a suppression mechanism is immediately activated to protect processing quality. The implementation of this step can be flexibly adjusted according to the actual situation of the processing equipment and is not subject to excessive limitations.

[0050] S1031. The transmitting end uses a vibration sensor to collect vibration frequency and amplitude data of the surface porous bone structure region. Time-frequency analysis is performed using Fourier transform to generate a vibration frequency spectrum, and the dominant frequency and harmonic components are extracted. Simultaneously, a Kalman filter is used to smooth the vibration amplitude data to generate a vibration amplitude curve. In this embodiment, the vibration sensor operates at a sampling frequency of 10 kHz, continuously collecting 10,000 data points per second, covering the vibration characteristics of the surface region. Fourier transform converts the vibration frequency data into a frequency spectrum, revealing that the dominant frequency is concentrated at 156 Hz, accompanied by harmonic components at 312 Hz and 468 Hz, reflecting the periodicity of the vibration. The Kalman filter eliminates noise through iterative estimation, making the vibration amplitude curve smoother, with the measured amplitude fluctuating between 0.12 mm and 0.28 mm. This processing method effectively separates the main trend and noise in the vibration signal, providing a clear data foundation for subsequent analysis.

[0051] S1032. The delivery end calculates the vibration frequency change rate based on the vibration frequency spectrum, constructs a mapping model between the vibration frequency change rate and porosity change using a recurrent neural network, and determines whether the vibration exhibits nonlinear characteristics and exceeds the safety threshold by combining the fluctuation range of the vibration amplitude curve.

[0052] In this embodiment of the invention, the vibration frequency change rate is calculated through the dynamic changes of the dominant frequency and its harmonic components. Within a 0.2-second detection window, the maximum change rate reaches 12 Hz per second, the minimum is -8 Hz per second, and the range is 20 Hz per second. The recurrent neural network adopts a three-layer fully connected hidden layer structure. The input layer receives frequency change rate data for 20 time steps, with the number of hidden layer neurons being 32, 16, and 8 respectively. The output layer predicts the porosity change value. After training, the model shows that the correlation between the frequency change rate and porosity is as high as 0.85, and the porosity fluctuation range is from 12% to 18%. If the preset nonlinearity judgment threshold is 15 Hz per second, the current change range exceeds the threshold, indicating that the vibration has nonlinear characteristics. At the same time, the vibration amplitude fluctuation range is 0.16 mm, exceeding the preset safety threshold of 0.15 mm, requiring further processing.

[0053] S1033 When abnormal vibration is detected at the feeding end, the proportional-integral controller dynamically adjusts the spindle speed and feed rate according to the suppression parameter mapping table until the vibration amplitude drops below the safety threshold. At the same time, the porosity fluctuation is monitored in real time to verify the suppression effect.

[0054] In this embodiment of the invention, the suppression parameter mapping table is pre-constructed based on the vibration frequency change rate and amplitude fluctuation range. For example, when the frequency change rate is between 15 and 20 Hz per second and the amplitude fluctuation is between 0.15 and 0.20 mm, it is recommended to reduce the spindle speed by 200 rpm and the feed rate by 50 mm per minute. The proportional-integral controller takes the amplitude deviation as input, with a proportional coefficient of 2 and an integral coefficient of 0.5, and outputs an adjustment to reduce the spindle speed from 3000 rpm to 2800 rpm and the feed rate from 200 mm per minute to 150 mm per minute. After adjustment, the frequency change rate is reduced to below 10 Hz per second, the amplitude fluctuation is reduced to 0.08 mm, and the porosity change is reduced from 6% to 3%, indicating a significant improvement in machining stability. The entire adjustment process takes approximately 0.5 seconds, ensuring a rapid response.

[0055] S1034. The feeding end verifies the nonlinear characteristics of the vibration parameter trend through multi-cycle data, uses a mapping model to predict the potential impact of porosity changes on vibration, and optimizes the triggering conditions of the suppression mechanism to adapt to different processing scenarios.

[0056] In this embodiment of the invention, the delivery end repeatedly collects data within multiple 0.2-second detection cycles, revealing a strong correlation between the nonlinear characteristics of the vibration frequency change rate and porosity fluctuations, particularly evident when the cutting force in the surface region is unstable. The mapping model predicts that when porosity increases by more than 5%, the vibration amplitude may increase by more than 0.1 mm. Based on this, the safety threshold is optimized to 0.14 mm, triggering the suppression mechanism in advance. This predictive capability enhances the system's adaptability to dynamic processing environments, ensuring the effectiveness of the suppression mechanism under different material properties.

[0057] In this embodiment of the invention, the delivery end achieves accurate judgment and timely suppression of vibration trends in the surface area through the aforementioned analysis and adjustment steps. The identification of nonlinear characteristics benefits from time-frequency analysis and neural network modeling, while the suppression process rapidly stabilizes vibration parameters through a precise control strategy. The advantage of this method lies in balancing processing efficiency and quality; subsequent adjustments to model parameters or control logic can be made to further improve performance based on actual needs.

[0058] S104. The feeding end extracts vibration parameters of the deep, dense bone structure region during continuous machining. By analyzing the coordinated changes in these parameters with the cutting state, it determines whether there is a positive correlation between vibration frequency and cutting force. After confirming the correlation, the parameter change patterns are recorded to construct a mapping table, providing a basis for subsequent vibration suppression. In this embodiment of the invention, the feeding end uses multiple sensors to collaboratively collect vibration and cutting data from the deep region, and uses signal decomposition and time-series modeling methods to reveal the dynamic correlation between parameters. The analysis process aims to capture the stability of the machining state and its specific impact on vibration, ensuring that the recorded data can truly reflect the machining characteristics of the deep structure. The specific implementation of this step can be adjusted according to equipment performance and machining requirements.

[0059] S1041. The feeding end uses vibration sensors and cutting force sensors to collect vibration frequency and cutting force data of deep dense bone structure region in real time. The vibration frequency signal is decomposed by wavelet transform to extract frequency component and amplitude component, and the statistical characteristics of cutting force are calculated to determine the stability of processing state.

[0060] In this embodiment of the invention, the vibration sensor measured a vibration frequency range of 80 to 120 Hz in the deep region, while the cutting force sensor recorded a cutting force fluctuating between 200 and 300 Newtons. Wavelet transform employed a four-level decomposition method, with high-frequency components highlighting instantaneous vibration characteristics, such as the spike signal during tool entry, while low-frequency components reflected the smooth changes in the machining trend. The mean of the cutting force data was 250 Newtons, with a standard deviation of 20 Newtons, representing 8%, which is below the preset 10% threshold, indicating that the machining was in a stable range. At this time, the recorded vibration frequency mean was 100 Hz, with an amplitude of 0.08 mm, providing reliable data for subsequent analysis.

[0061] S1042. The delivery end uses Pearson correlation coefficient to analyze the correlation between vibration frequency and cutting force, combines depth sensor to measure the density distribution of deep bone structure, and constructs a time-series mapping model of cutting force and vibration frequency through long short-term memory network to extract cooperative change feature values.

[0062] In this embodiment of the invention, the Pearson correlation coefficient is calculated to be 0.82, indicating a strong positive correlation between vibration frequency and cutting force. The depth sensor, using ultrasonic technology, measures that when the depth is between 10 and 15 mm, the density increases from 88% to 96%, forming a density distribution curve that reflects the gradual changes in the deep structure.

[0063] The Long Short-Term Memory (LSTM) network takes the time sequence of cutting force and vibration frequency as input. The network contains two hidden layers, each with 50 units, and outputs co-variation feature values. For example, when the cutting force increment is 50 Newtons and the vibration frequency increment is 15 Hz, the feature value reaches 0.75, indicating a significant co-variation between the two. This model can remember long-term dependencies in the machining process, improving the accuracy of feature extraction.

[0064] S1043. The feeding end determines the positive correlation state based on the tool wear data and the cooperative change characteristic value. If the ratio of the vibration frequency increment to the cutting force increment exceeds the preset threshold, the relevant parameters are recorded and a cutting state parameter mapping table is constructed to quantify the dynamic law of the machining process.

[0065] In this embodiment of the invention, the tool wear sensor detects that the wear amount increases from 0.1 mm to 0.3 mm. When the wear amount reaches 0.2 mm, the cutting force increases to 280 N, and the vibration frequency rises to 110 Hz. The increment ratio is 0.35 Hz per N, which is higher than the preset threshold of 0.3 Hz per N, indicating a strong positive correlation. The recorded parameters at this time include a vibration frequency of 110 Hz, an amplitude of 0.09 mm, a cutting force of 280 N, and a wear amount of 0.25 mm. The mapping table further records the parameter combinations at different stages. For example, when the wear amount is 0.15 mm, the vibration frequency is 95 Hz, the amplitude is 0.06 mm, and the cutting force is 220 N; when the wear amount increases to 0.28 mm, the parameters become 115 Hz, 0.1 mm, and 300 N. This mapping relationship clearly shows the evolution path of the machining parameters.

[0066] S1044. The feeding end verifies the stability of vibration parameters by testing the physical characteristics of the deep region, analyzes the potential impact of cutting temperature and elastic modulus on vibration using multidimensional data, and optimizes the application range of the mapping table to adapt to complex machining conditions.

[0067] In this embodiment of the invention, the elastic modulus of the deep region was measured to be 18 GPa, significantly higher than the 5 GPa of the surface region, resulting in a more uniform stress distribution and enhanced vibration stability. Infrared thermography showed a cutting temperature of approximately 200 degrees Celsius, 150 degrees Celsius lower than the surface temperature, reducing the interference of thermal deformation on vibration. After optimization based on these physical properties, the mapping table maintains accurate parameter recording even with increased tool wear or changes in cutting depth. For example, when the depth increases to 15 mm and the density approaches 96%, the vibration amplitude is only 0.05 mm, reflecting the vibration-suppressing effect of high density. This multidimensional analysis provides broader applicability for the mapping table.

[0068] In this embodiment of the invention, the delivery end comprehensively extracted and analyzed the synergistic variation characteristics of vibration parameters and cutting state in the deep region through the above steps. The construction of the mapping table not only quantifies the dynamic laws of the machining process, but also lays the foundation for subsequent vibration trend prediction and suppression strategy formulation. The method makes full use of the signal processing capabilities of wavelet transform and long short-term memory networks to ensure the depth and accuracy of data analysis. The parameter dimensions in the table can be further enriched according to machining requirements.

[0069] S105. Based on the vibration parameter variation law of the deep dense bone structure region, the feeding end establishes a collaborative variation model in combination with cutting state data. Through multi-dimensional feature extraction and time-series prediction methods, the vibration trend of the spindle in the low porosity region is derived, providing a scientific basis for the optimization of the machining process.

[0070] In this embodiment of the invention, the delivery end collects vibration and cutting parameters of the deep region through multiple sensors, and constructs a predictive model using signal processing and machine learning techniques. The model aims to reveal the intrinsic relationship between vibration and cutting conditions, and to predict vibration behavior under low porosity conditions. The implementation of this step can be flexibly adjusted according to the processing equipment and material characteristics.

[0071] S1051. The delivery end uses a vibration sensor to collect the original vibration parameter data of the deep dense bone structure region, and converts it into a standard range through range normalization. At the same time, it combines a cutting state sensor and an ultrasonic porosity detector to obtain cutting force and porosity data, laying the foundation for subsequent feature extraction.

[0072] In this embodiment of the invention, the vibration sensor operates at a sampling frequency of 10 kHz, acquiring vibration frequencies ranging from 80 to 120 Hz, with amplitudes between 0.05 and 0.15 mm. Range normalization maps these data to the 0-1 interval, avoiding interference from dimensional differences in the analysis. The cutting force sensor records force values ​​fluctuating between 220 and 280 Newtons, and the ultrasonic testing instrument measures a stable porosity of 3% to 5% in the low-porosity region. These multi-source data collectively constitute a comprehensive characterization of the machining state, ensuring the richness and reliability of the model input.

[0073] S1052. The sending end performs a short-time Fourier transform on the normalized vibration parameter data to generate a time-frequency feature matrix, extracts the vibration frequency and amplitude features, and integrates the spindle speed and cutting force data to construct a multi-dimensional feature vector to capture the dynamic characteristics of the machining process.

[0074] In this embodiment of the invention, the short-time Fourier transform uses a 256-point window with a 50% overlap rate. The generated time-frequency feature matrix shows that the dominant frequency component is concentrated around 100 Hz, and the amplitude spectrum varies from 0.08 to 0.12 mm. The spindle speed sensor records a stable speed of 3000 rpm, and the cutting force data reflects the real-time changes in the machining load. The multidimensional feature vector integrates 8 frequency features, 8 amplitude features, as well as the speed and cutting force values, for a total of 18 dimensions. This high-dimensional representation method can comprehensively reflect the interaction between vibration and cutting conditions, providing solid data support for model training.

[0075] S1053. The feeding end constructs a collaborative change model of vibration parameters and cutting state through a recurrent neural network, and uses a Bayesian regressor combined with measured porosity data to optimize the prediction output and generate a vibration trend prediction curve to determine the correlation and change trend between vibration and cutting state.

[0076] In this embodiment of the invention, the recurrent neural network is designed as a three-layer structure. The input layer has 18 neurons that receive multidimensional feature vectors, the hidden layer contains 32 long short-term memory units to capture temporal dependencies, and the output layer has 4 neurons that predict changes in vibration frequency and amplitude. The model is trained based on 1000 sets of time-series samples, with 20 time steps per set. A Bayesian regressor optimizes the relationship between the predicted value and the measured porosity using a Gaussian prior distribution. The output curve shows that the vibration frequency fluctuation in the low porosity region is less than 5 Hz. For example, when the frequency increment is 10 Hz and the cutting force increment is 35 N, the ratio of 0.286 Hz / N exceeds the threshold of 0.25 Hz / N, indicating a significant positive correlation. At this time, the predicted frequency is 105 Hz and the amplitude is 0.09 mm, demonstrating the model's predictive ability.

[0077] S1054. The delivery end uses cubic spline interpolation to smooth the vibration trend prediction curve, calculates the trend slope in the low porosity region, and evaluates the model accuracy through multiple rounds of verification to ensure the reliability of the prediction results in engineering applications.

[0078] In this embodiment of the invention, cubic spline interpolation uses 20 control points to fit the predicted curve, and the slope of the low porosity region is calculated to be 0.15 Hz per second, indicating that the vibration change is relatively gentle. Model verification shows that the average prediction error is controlled within 5%, and when the porosity is below 4%, the frequency prediction error is less than 3%, and the amplitude prediction accuracy reaches 0.01 mm. This high accuracy stems from the model's deep fusion of multi-source data and accurate capture of processing physical characteristics, enabling the prediction results to effectively guide subsequent parameter adjustments.

[0079] In this embodiment of the invention, the feeding end successfully established a collaborative change model of deep region vibration and cutting state through the above steps, and achieved accurate prediction of vibration trends. The advantage of the method lies in combining the powerful processing capabilities of time-frequency analysis and neural networks, ensuring the controllability of the processing in low porosity regions, and the model can be further expanded to adapt to more complex processing scenarios.

[0080] S106. The feeding end extracts the vibration parameters to obtain key features such as vibration amplitude, frequency and rate of change. It also analyzes the vibration trend in conjunction with the collaborative change model. If it finds that the amplitude continues to increase while the frequency tends to stabilize, it triggers an automatic suppression mechanism to optimize the processing stability.

[0081] In this embodiment of the invention, the delivery end comprehensively extracts vibration features and determines their dynamic behavior through multi-scale signal processing and prediction models. The suppression mechanism is triggered based on the coordinated change of feature quantities, ensuring timely control of vibration during processing. The implementation details of this step can be flexibly optimized according to actual processing conditions.

[0082] S1061. The transmitting end uses a vibration sensor to collect vibration data sequences, and uses wavelet transform to perform multi-scale decomposition to extract vibration amplitude and frequency data. After smoothing the amplitude data with a Kalman filter, its rate of change is calculated to provide accurate input for subsequent trend analysis.

[0083] In this embodiment of the invention, the vibration sensor operates at a sampling frequency of 10 kHz, collecting 1000 data points per cycle. Wavelet transform is performed using the db4 basis function, decomposed into three levels. High-frequency coefficients reflect amplitude changes, ranging from 0.05 to 0.15 mm, while low-frequency coefficients extract frequency features, fluctuating between 80 and 120 Hz. The Kalman filter is set with a state noise covariance of 0.01 and a measurement noise of 0.1. After filtering, the amplitude curve is smoother, and the calculated rate of change within a 200-millisecond window is 0.002 mm / s. This processing method effectively reduces noise interference and improves the accuracy of feature extraction.

[0084] S1062. The transmitting end constructs a vibration parameter predictor through a recurrent neural network, inputs the vibration frequency, amplitude and rate of change, predicts the future trend, and combines the frequency variance and amplitude rate of change to determine whether to trigger the suppression controller. At the same time, it generates a suppression command based on the prediction results.

[0085] In this embodiment of the invention, the recurrent neural network comprises three layers: the input layer receives three features—frequency, amplitude, and rate of change; the hidden layer with 32 recurrent units captures temporal relationships; and the output layer predicts the vibration trend at the next five time points. Within a 200-millisecond window, the frequency variance is 25 square Hz, below the stability threshold of 30 square Hz; the amplitude change rate is 0.0015 mm / s, above the growth threshold of 0.001 mm / s, indicating amplitude growth and frequency stability. The prediction results show a continuous upward trend in amplitude, triggering the suppression controller. The controller looks up a table to determine the rotational speed adjustment as -200 rpm and the feed rate adjustment as -50 mm / s, generating and sending commands through the CNC system.

[0086] S1063. The feeding end executes the suppression command, uses a proportional-integral controller to realize closed-loop regulation of rotational speed and feed rate, monitors vibration data in real time to verify the suppression effect, and dynamically adjusts control parameters to adapt to changes in processing conditions.

[0087] In this embodiment of the invention, the proportional-integral controller has a proportional coefficient of 2.5 and an integral coefficient of 0.5. The output gain is linearly adjusted with the rate of change, for example, a gain of 1.5 when the rate of change is 0.0015 mm / s. The actuator reduces the rotational speed from 3000 rpm to 2800 rpm and the feed rate from 200 mm / min to 150 mm / min. Real-time data shows that after suppression, the rate of change drops to below 0.0008 mm / s in approximately 500 milliseconds, and the frequency variance remains within 20 square Hz. Tests show that when the workpiece hardness increases from 35 HRC to 45 HRC or the tool wear reaches 0.3 mm, the controller can still adaptively adjust the parameters to ensure stable suppression effect.

[0088] S1064. The delivery end optimizes the suppression mechanism through multi-condition verification, and uses the cooperative change characteristics to analyze the coupling relationship between vibration and processing status, thereby improving the adaptability of the control strategy to complex environments and ensuring processing efficiency.

[0089] In this embodiment of the invention, repeated tests were conducted on the feeding end under different hardness and tool wear conditions. It was found that the synergistic characteristics of the amplitude change rate and frequency variance accurately reflect changes in the machining state. For example, when the hardness increases, the amplitude change rate briefly rises to 0.002 mm / s, and the controller quickly adjusts the rotational speed and feed rate to restore them below the threshold. This adaptive capability stems from the comprehensiveness of feature extraction and the robustness of the prediction model, ensuring both machining efficiency and vibration optimization.

[0090] In this embodiment of the invention, the delivery end achieves accurate judgment of vibration trends through feature extraction and predictive analysis, and effectively suppresses abnormal vibrations through closed-loop control. The advantage of this method lies in combining the efficiency of wavelet transform and neural networks, enabling rapid response in dynamic processing environments. Furthermore, the control parameters can be further refined according to actual needs to expand its applicability.

[0091] S107. The feeding end adjusts the spindle speed and feed rate through a preset automatic suppression mechanism to reduce the vibration amplitude and stabilize the frequency. At the same time, it monitors the cutting status in real time and re-acquires vibration parameters. It judges whether the suppression effect has reached the expected level through feature extraction and mapping analysis. If the amplitude drops below the threshold and the frequency is stable, the suppression process is completed.

[0092] In this embodiment of the invention, the feeding end utilizes multi-sensor data and neural network technology to achieve dynamic vibration suppression. The suppression process not only focuses on real-time changes in vibration parameters but also incorporates feedback from the cutting state to ensure continuous optimization of machining quality. The implementation method of this step can be flexibly adjusted according to actual working conditions.

[0093] S1071. The feeding end uses a suppression parameter selector to obtain the speed and feed adjustment amount from the preset mapping table, sends the command to the spindle and feed controller through the CNC communication interface, and uses the proportional-integral controller to realize closed-loop adjustment based on the feedback signal.

[0094] In this embodiment of the invention, the mapping table is constructed based on experimental data. When the vibration amplitude is between 0.1 and 0.15 mm, the recommended speed adjustment is -200 rpm and the feed adjustment is -50 mm / min. Commands are sent via industrial bus at 100 millisecond cycles. The spindle encoder has a resolution of 2500 pulses per revolution, and the feed encoder has a resolution of 0.1 micrometers per pulse, providing high-precision feedback. The proportional-integral controller initially has a proportional coefficient of 2.5 and an integral coefficient of 0.5. When the amplitude exceeds 0.12 mm, the proportional coefficient increases to 3.0 and the integral coefficient increases to 0.8, ensuring the speed and stability of adjustment.

[0095] S1072. The feeding end collects vibration data during the adjustment process through vibration sensors, smooths the data using a Kalman filter, inputs it into a recurrent neural network to extract features, and combines the cutting force data to establish a mapping relationship between vibration and cutting state to determine the stable range.

[0096] In this embodiment of the invention, the vibration sensor collects data at a sampling frequency of 10 kHz. The initial value of the state noise covariance of the Kalman filter is 0.01, which is adjusted to a maximum of 0.05 as the data fluctuates. The measurement noise covariance varies from 0.1 to 0.2. The input layer of the recurrent neural network contains 20 neurons, receiving vibration sequences within a 200-millisecond window. The hidden layer has 32 long short-term memory units, with a forget gate threshold of 0.6, and outputs frequency stability and amplitude change rate. The cutting force sensor measures force values ​​between 200 and 300 Newtons. The mapping relationship is generated through cubic spline interpolation with an interpolation interval of 50 Newtons. The stable range is a frequency of 90 to 110 Hz and an amplitude of 0.05 to 0.08 mm.

[0097] S1073. The feeding end judges the suppression effect based on the vibration characteristics and stable range. If the amplitude is less than 0.1 mm and the frequency stability is greater than 0.8, the parameters are kept unchanged. Otherwise, the adjustment amount is recalculated and the suppression process is repeated. At the same time, the adaptability of the mechanism to hardness and tool wear is verified.

[0098] In this embodiment of the invention, when the amplitude drops to 0.08 mm and the stability reaches 0.85, the rotational speed remains at 2800 rpm, the feed rate at 150 mm / min, and the cutting force stabilizes at 250 N. If the amplitude suddenly increases to 0.13 mm, the system calculates a new adjustment, reducing the rotational speed by another 250 rpm and the feed rate by 60 mm / min, and stabilizing after 3 to 4 cycles. Tests show that when the hardness increases from 35 HRC to 45 HRC or the tool wear reaches 0.3 mm, the controller adaptively adjusts the coefficients to maintain the suppression effect, demonstrating strong robustness.

[0099] S1074. The feeding end uses real-time monitoring and repeated adjustment to optimize the suppression mechanism, and uses multi-dimensional data analysis to analyze the potential impact of changes in cutting state on vibration, ensuring the continuous stability and efficiency of machining parameters in dynamic environments.

[0100] In this embodiment of the invention, the feeding end continuously collects vibration and cutting data. It detects a brief increase in amplitude when hardness increases, and the controller quickly adjusts the parameters to bring it back below the threshold. When tool wear intensifies, the mapping relationship shows a slight shift in the frequency stability range. The system adapts to the change by increasing the adjustment amount, and the entire process converges rapidly with a response time of approximately 300 to 500 milliseconds. This dynamic optimization capability ensures both machining efficiency and vibration control, and the mapping table can be further refined based on the operating conditions.

[0101] The above description is only a preferred embodiment of the present invention. It should be noted that those skilled in the art can make several improvements and additions without departing from the principle of the present invention, and these improvements and additions should also be considered within the scope of protection of the present invention.

Claims

1. A method for real-time monitoring and automatic suppression of spindle vibration in CNC machine tools, characterized in that, The method includes: The spindle vibration parameters of the biomimetic structure of the femoral stem of the titanium alloy artificial hip joint were obtained during the CNC machine tool machining process. The vibration parameters included vibration frequency and vibration amplitude. At the same time, the porosity distribution information of the machining area was collected. Based on the porosity distribution information, the processing area is divided into a surface loose bone structure area and a deep dense bone structure area. Vibration parameters during the processing of different areas are extracted, and a mapping relationship between vibration parameters and porosity distribution is established. Analyze the vibration parameter variation trend in the surface porous bone structure region, and determine whether the vibration parameter exhibits nonlinear fluctuation with the change of porosity based on the vibration parameter variation trend. If so, and the vibration amplitude exceeds the preset safety threshold, the preset vibration suppression mechanism is immediately triggered. Vibration parameters of deep dense bone structure regions during continuous machining are extracted to determine whether the vibration parameters exhibit coordinated changes with the cutting state. If the vibration frequency is positively correlated with the cutting force, the variation law of vibration parameters under the cutting state is recorded. Based on the vibration parameter variation law of the deep dense bone structure region, a coordinated variation model of vibration parameters and cutting state is established to predict the vibration trend of the spindle when machining in the low porosity region, including: The original vibration parameter data is obtained by using a vibration sensor, and the original vibration parameter data is normalized to obtain normalized vibration parameter data by the range normalization method. Calculate the short-time Fourier transform based on the normalized vibration parameter data to obtain the time-frequency feature matrix, and extract the vibration frequency features and vibration amplitude features from the time-frequency feature matrix; Spindle speed sensor and cutting force sensor acquire speed data and cutting force data, and construct multidimensional feature vector based on the vibration frequency characteristics, vibration amplitude characteristics, speed data and cutting force data; A recurrent neural network is used to establish a cooperative change model. The input layer of the cooperative change model is the multidimensional feature vector. The output value of the cooperative change model and the measured porosity value are used to train a Bayesian regressor. The Bayesian regressor outputs a vibration trend prediction curve. Vibration parameters are feature extracted to extract key features such as vibration amplitude, vibration frequency, and the rate of change of vibration parameters over time. Combined with the cooperative change model, it is determined whether the vibration amplitude continues to increase and the vibration frequency tends to stabilize. If so, a preset automatic suppression mechanism is triggered. The automatic suppression mechanism adjusts the spindle speed and feed rate to reduce vibration amplitude and stabilize vibration frequency. At the same time, it monitors the changes in cutting status in real time, re-collects vibration parameters, and judges whether the vibration suppression effect meets the preset requirements. If the vibration amplitude drops below the threshold and the vibration frequency tends to be stable, the automatic suppression process is completed.

2. The method according to claim 1, characterized in that, The process of acquiring spindle vibration parameters during CNC machine tool machining of the biomimetic femoral stem structure of the titanium alloy artificial hip joint includes vibration frequency and vibration amplitude. Simultaneously, porosity distribution information of the machining area is collected, including: The spindle rotation frequency data is acquired from the spindle speed sensor and the vibration frequency data is acquired from the vibration sensor. A parameter correlation matrix is ​​established based on the spindle rotation frequency data and the vibration frequency data. A vibration parameter distribution map is established based on the parameter correlation matrix, and the vibration parameter distribution map is used to obtain porosity distribution data of the processing area. Acoustic emission signals are collected using acoustic emission sensors, and a mapping relationship is established between the acoustic emission signals and the porosity distribution data of the processing area. Determine whether the acoustic emission signal exceeds a preset threshold range. If the acoustic emission signal exceeds the preset threshold range, adjust the tool feed rate and cutting depth until the acoustic emission signal drops to within the preset threshold range.

3. The method according to claim 1, characterized in that, Based on the porosity distribution information, the processing area is divided into a surface porous bone structure region and a deep dense bone structure region. Vibration parameters during the processing of different regions are extracted, and a mapping relationship between the vibration parameters and the porosity distribution is established, including: Density clustering is performed on the point cloud dataset. Based on the porosity distribution density, the point sets of loosely structured regions and the point sets of densely structured regions are marked. Boundary labeling results are obtained by using region boundary determination rules. The width of the structural transition zone between the surface porous bone structure region and the deep dense bone structure region is calculated based on the boundary calibration results. Vibration parameters within the surface region are collected using a vibration sensor, and the vibration parameters are normalized to obtain the surface vibration feature vector. The depth parameters of the bone structure in the point cloud dataset are measured by a depth detector. Support vector regression is used to establish the mapping relationship between the vibration feature vector and porosity. The vibration parameters in the deep region are normalized to obtain the deep vibration feature vector. A mapping function is constructed based on the surface vibration feature vector and the deep vibration feature vector. A porosity threshold is extracted from the mapping function. If the porosity of the detection point is higher than the porosity threshold, it is classified into the surface loose bone structure region. If the porosity of the detection point is lower than the porosity threshold, it is classified into the deep dense bone structure region.

4. The method according to claim 1, characterized in that, The analysis examines the vibration parameter variation trend in the surface porous bone structure region. Based on this trend, it determines whether the vibration parameters exhibit nonlinear fluctuations with changes in porosity. If so, and the vibration amplitude exceeds a pre-set safety threshold, a preset vibration suppression mechanism is immediately triggered, including: Vibration frequency data and vibration amplitude data of the surface porous bone structure region are acquired using vibration sensors. The vibration frequency data are then analyzed in time and frequency using Fourier transform to obtain the vibration frequency spectrum. The vibration frequency change rate is calculated based on the dominant frequency component and harmonic component in the vibration frequency spectrum. The vibration amplitude data is then filtered using a Kalman filter to obtain the vibration amplitude curve. A recurrent neural network mapping model is established for the vibration frequency change rate sequence. The recurrent neural network includes an input layer and three fully connected hidden layers and an output layer. The porosity change value is obtained from the mapping model. If the rate of change of vibration frequency and the fluctuation range of vibration amplitude exceed the preset safety threshold, the spindle speed and feed speed are dynamically adjusted according to the proportional-integral controller until the vibration amplitude drops below the safety threshold.

5. The method according to claim 1, characterized in that, The vibration parameters of the deep, dense bone structure region are extracted during continuous machining. It is determined whether the vibration parameters exhibit a coordinated change characteristic with the cutting state. If the vibration frequency is positively correlated with the cutting force, the variation law of the vibration parameters under this cutting state is recorded, including: Vibration frequency data and cutting force data are acquired using vibration sensors and cutting force sensors. The vibration frequency data is then decomposed using wavelet transform to obtain vibration frequency components and vibration amplitude components. The mean and standard deviation of the cutting force are calculated based on the cutting force data. If the standard deviation is less than the preset ratio of the mean, it is determined to be a stable cutting state interval, and the vibration frequency component and the vibration amplitude component are recorded. The Pearson correlation coefficient is calculated based on the vibration frequency components and the cutting force data, and the fluctuation value of the vibration parameters of the deep bone structure is obtained from the correlation coefficient. The cutting force data and the vibration frequency data are processed using a long short-term memory network to obtain a co-variation feature value. If the co-variation feature value is greater than a preset threshold, the vibration frequency component and the vibration amplitude component are recorded with the cutting force data to construct a cutting state parameter mapping table.

6. The method according to claim 1, characterized in that, The vibration parameters are feature extracted, including vibration amplitude, vibration frequency, and the rate of change of vibration parameters over time. Combined with a collaborative change model, it is determined whether the vibration amplitude continues to increase and the vibration frequency tends to stabilize. If so, a preset automatic suppression mechanism is triggered, including: Vibration data sequences are acquired using vibration sensors, and the vibration data sequences are decomposed into multi-scale values ​​and vibration frequency data by wavelet transform. The frequency variance within the time window is calculated based on the vibration frequency data, the vibration amplitude data is filtered using a Kalman filter, and the amplitude change rate is calculated from the filtered vibration amplitude data. The vibration frequency data, vibration amplitude data, and amplitude change rate are processed by a recurrent neural network to obtain vibration parameter prediction results. If the frequency variance is less than the stability threshold and the amplitude change rate is greater than the growth threshold, the suppression controller is triggered. The suppression controller selects the spindle speed adjustment amount and the feed speed adjustment amount according to the vibration parameter prediction result and generates a suppression control command.

7. The method according to claim 1, characterized in that, The process involves adjusting the spindle speed and feed rate through a preset automatic suppression mechanism to reduce vibration amplitude and stabilize vibration frequency. Simultaneously, it monitors changes in the cutting state in real time, re-collects vibration parameters, and determines whether the vibration suppression effect meets preset requirements. If the vibration amplitude drops below a threshold and the vibration frequency tends to stabilize, the automatic suppression process is complete, including: Vibration sensors are used to acquire vibration frequency and amplitude data during the cutting process. The vibration data is then filtered by a Kalman filter to obtain filtered vibration data. Vibration features are extracted from the filtered vibration data using a recurrent neural network, and the recurrent neural network outputs the vibration frequency stability and the vibration amplitude change rate. Cutting force data during the cutting process is collected by a cutting force sensor, and a mapping relationship is established between the vibration characteristics and the cutting force data to obtain the stable range of vibration parameters; If the vibration frequency stability is lower than the preset stability threshold, the speed adjustment and feed adjustment are calculated based on the vibration parameters, and the adjustment is sent to the spindle controller and feed controller through the CNC communication interface.

Images