Milling process vibration state monitoring method and system based on acceleration signal

Abstract

In order to solve the technical problems that the existing method for identifying the vibration state by establishing a dynamic model can not guarantee the reliability of the identification effect, and the method for establishing an identification model based on artificial intelligence technology can only identify a single index and is prone to missing reports and false reports, the present application provides a milling process vibration state monitoring method based on an acceleration signal. After the unstable vibration state is identified by the standard deviation std of the energy strongest intrinsic modal component, the ratio stdr of the standard deviations of the milling signal x(n) and the chatter related component e(n) is further combined to distinguish the two unstable vibration states, so that the vibration state categories covered are more comprehensive. Since two identification indexes are used to quickly identify the occurrence of unstable vibration, the severe vibration and the chatter can be accurately distinguished, and missing reports and false reports can be avoided.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of mechanical processing state detection, and particularly relates to a milling process vibration state monitoring method and system. BACKGROUND

[0002] When milling thin-walled structural parts, due to the low strength and stiffness characteristics, improper cutting depth and spindle speed and other cutting parameters can cause unstable vibration states of the process system, mainly including two kinds of severe vibration and chatter. The former is due to excessive milling force, which causes the amplitude of forced vibration of the workpiece to increase; and the latter is a self-excited vibration caused by dynamic changes in cutting thickness. Unstable vibration state is the main reason for causing rough surface quality of the workpiece, size precision out of tolerance and even damage to the process system. Therefore, how to accurately and reliably identify the current vibration state and timely warn the unstable vibration state is one of the key means to avoid abnormal vibration, improve processing efficiency and ensure processing quality.

[0003] The main idea of the traditional vibration state identification and prediction method is to establish a dynamic model of the process system, solve the relationship between the cutting depth, spindle speed and system state, and draw the corresponding stability domain lobe diagram, so as to realize the prediction and identification of the vibration state. However, since the established dynamic model ignores many factors such as workpiece material and machine tool structure, and the time-varying nature caused by factors such as tool wear and temperature change, the stability domain lobe diagram obtained does not match the actual situation, and it is difficult to ensure the reliability of the vibration state identification effect. Therefore, the method of real-time monitoring of vibration state is an effective means to warn and avoid unstable vibration state. In order to realize online monitoring of vibration state, some studies use artificial intelligence technology to automatically establish an identification model. However, artificial intelligence technology needs a large number of experimental samples for training, and has poor applicability and transferability. In addition, from the existing search literature, it is found that at present, a large number of monitoring methods are based on single index identification, which cannot accurately reflect the vibration state of the system, is prone to false negatives and false positives, and has poor robustness.

[0004] Therefore, it is necessary to propose an identification index reflecting the vibration state and having small calculation amount, to accurately monitor the vibration state, establish a milling process vibration state monitoring method, solve the problems of limited applicability and poor transferability, and have important significance for online monitoring of vibration state. SUMMARY

[0005] In order to solve the technical problems that the existing method for identifying the vibration state by establishing a dynamic model cannot guarantee the reliability of the identification effect, and the method for establishing an identification model based on artificial intelligence technology can only identify a single index and is prone to missing reports and false reports, the application provides a milling process vibration state monitoring method based on an acceleration signal.

[0006] The technical solution adopted by the application to solve the technical problems is:

[0007] The milling process vibration state monitoring method based on an acceleration signal is characterized in that it comprises the following steps:

[0008] Step 1: Collecting an acceleration signal during milling and denoising to obtain a pure milling signal;

[0009] Step 2: Empirical mode decomposition of the milling signal to obtain a series of intrinsic mode components, calculation of the standard deviation of the intrinsic mode component with the strongest energy, comparison of the standard deviation with a preset threshold value one, if the standard deviation is less than the threshold value one, the current state is a stable forced vibration state, and returning to step 1; if the standard deviation is greater than or equal to the threshold value one, the current state is an unstable vibration state, and entering step 3;

[0010] Step 3: Extraction of a chatter-related component from the milling signal;

[0011] Step 4: Calculation of the ratio of the amplitude standard deviation of the chatter-related component to the amplitude standard deviation of the milling signal, comparison of the ratio with a preset threshold value two, if the ratio is less than the threshold value two, the current state is a severe vibration state; if the ratio is greater than or equal to the threshold value two, the current state is a chatter state.

[0012] Further, the denoising method in step 1 is based on a non-cutting signal segment to estimate noise and eliminate noise in the acceleration signal to obtain a pure milling signal.

[0013] Further, the denoising method in step 1 is specifically:

[0014] First, the acceleration signal is converted into a frequency domain signal, then the average amplitude spectrum of the frequency domain signal and the amplitude spectrum of the noise signal contained therein are calculated using a non-cutting signal segment, and finally the amplitude spectrum of the milling signal is obtained by operating the average amplitude spectrum and the amplitude spectrum of the noise signal and inverse transforming it to the time domain to obtain a pure milling signal.

[0015] Furthermore, the method for empirical mode decomposition of the milling signal in step 2 to obtain a series of intrinsic mode components is as follows:

[0016] 2.1) Find all local maximum and local minimum points of the milling signal x(n), and use cubic spline interpolation to connect the local maximum and minimum points to form the upper envelope e. max (n) and lower envelope e min (n);

[0017] 2.2) Calculate the upper envelope e max (n) and lower envelope e min (n) The average value at each time point yields the mean envelope.

[0018] 2.3) Subtract the mean envelope from the milling signal x(n) Once the first component h1(n) is obtained, determine whether the first component h1(n) satisfies the condition of an intrinsic mode component. If it does, then the first component h1(n) is the first-order intrinsic mode component IMF1(n). If not, proceed to step 2.4). If it satisfies the condition, proceed to step 2.5.

[0019] 2.4) Decompose the components h1(n) and find the upper envelope h of h1(n). 1max (n), lower envelope h 1min (n) and mean envelope Obtain component h 11 (n) and determine h 11 (n) Check if the conditions for intrinsic modal components are met. If not, repeat this step until component h is found. 1k (n) satisfies the condition for intrinsic modal components, at which point component h 1k (n) represents the first-order intrinsic mode component IMF1(n);

[0020] 2.5) Subtract the first-order intrinsic mode component IMF1(n) from the milling signal x(n) to obtain the margin r1(n);

[0021] 2.6) Using the residual r1(n) as the new original signal, repeat steps 2.1)-2.5) until the l-th intrinsic mode component (IMF) is reached. l (n) corresponds to the remaining amount r l The decomposition ends when (n) is a monotonic function or a constant.

[0022] The conditions for the intrinsic mode components are: 1) the number of zeros crossed by the component is equal to the number of extreme points over the entire time range, or the difference is at most 1; 2) at any time, the average value of the upper envelope determined by the local maximum point and the lower envelope determined by the local minimum point is 0.

[0023] Further, in the step 3, the milling signal is separated into a forced vibration component and a chatter-related component by using a filter to realize extraction of the chatter-related component.

[0024] Further, the filter is a least mean square (LMS) filter with an adaptive filtering function.

[0025] The application further provides a milling process vibration state monitoring system based on an acceleration signal, and the speciality thereof lies in comprising:

[0026] A signal acquisition and preprocessing unit is configured to acquire an acceleration signal during milling and perform noise reduction processing on the acceleration signal to obtain a pure milling signal.

[0027] A vibration state stability determination unit is configured to perform empirical mode decomposition on the milling signal to obtain a series of intrinsic mode components, calculate a standard deviation of an intrinsic mode component with the strongest energy as a vibration state stability evaluation index, compare the evaluation index with a preset threshold, and determine whether the current vibration state is stable according to a comparison result.

[0028] A chatter component extraction unit is configured to extract a chatter-related component from the milling signal.

[0029] A vibration state recognition unit is configured to calculate a ratio of an amplitude standard deviation of the chatter-related component to an amplitude standard deviation of the milling signal, compare the ratio with a preset threshold as a state recognition index, and recognize the current unstable state as a chatter or a severe vibration state according to a comparison result.

[0030] Further, the signal acquisition and preprocessing unit comprises an acceleration sensor and a data acquisition device connected to the acceleration sensor, the acceleration sensor is installed in a normal direction of a surface to be machined of a workpiece, and the chatter component extraction unit comprises a filter, and the filter is an adaptive filter.

[0031] The application further provides a non-volatile computer readable storage medium, and the computer readable storage medium stores a computer program, and the speciality thereof lies in that the computer program is used to execute the milling process vibration state monitoring method when being executed.

[0032] The application further provides an electronic device, and the electronic device comprises a processor and a storage medium, and the storage medium stores a computer program, and the speciality thereof lies in that the computer program is used to execute the milling process vibration state monitoring method when being executed by the processor.

[0033] The application has the following beneficial effects:

[0034] 1. The design principle of the identification index in the application is based on the vibration characteristics of unstable state, which can quickly identify the unstable vibration state and accurately distinguish between violent vibration and chatter, thereby guiding targeted vibration suppression.

[0035] 2. The identification index of the unstable state proposed in the application has a clear physical meaning. By decomposing the milling signal to obtain the eigenvibration component with the strongest energy, the sensitivity of monitoring is effectively improved.

[0036] 3. The unstable vibration state that can be identified by the application includes two kinds of violent vibration state and chatter state. After the unstable vibration state is identified by the standard deviation std of the eigenvibration component with the strongest energy, the ratio stdr of the standard deviations of the milling signal x(n) and the chatter-related component e(n) is further used to distinguish the two kinds of unstable vibration state, thereby comprehensively covering the vibration state categories.

[0037] 4. The identification index of the chatter state proposed in the application can reflect the essential physical characteristics of chatter. By filtering to separate the component reflecting chatter and the component irrelevant to chatter, the accuracy of chatter identification is effectively improved.

[0038] 5. The two identification indexes proposed in the application have small calculation amount and can be quickly applied to different milling conditions through simple experiments, thereby having wide application range and good migratability.

[0039] 6. In the application, the input of the filter is the spindle rotation frequency and the tooth passing frequency and their multiples, which can ensure that the obtained chatter component is more accurate, thereby ensuring the accuracy of chatter identification.

[0040] 7. The application eliminates the interference of noise by performing noise reduction processing on the collected signal, so that the obtained milling signal can more effectively reflect the vibration state, thereby improving the reliability of vibration state monitoring.

[0041] 8. Considering the condition limitations of the processing site, it may not be possible to support the installation of multiple sensors. Therefore, the application installs the acceleration sensor in the normal direction of the surface to be milled, where the captured vibration response is most significant. Therefore, the purpose of collecting signals reflecting the workpiece vibration condition through only one acceleration sensor can be achieved.

[0042] 9. Considering that the collected signal has non-stationary and nonlinear characteristics, the application further adopts a least mean square (LMS) filter for filtering. The filter has adaptive characteristics and can be optimized to the best filtering effect without human intervention.

[0043] 10. The application can realize real-time online monitoring of the vibration state during the milling process. BRIEF DESCRIPTION OF DRAWINGS

[0044] Figure 1 Flow chart of the vibration state monitoring method of the present application.

[0045] Figure 2 Principle diagram of the vibration state monitoring system of the present application.

[0046] Figure 3 Structure diagram of the adaptive filter in the present application.

[0047] Figure 4 Example diagram of the severe vibration state in the present application, wherein (a) is a milling signal of the severe vibration state, (b) is a spectrum diagram of the milling signal in (a), (c) is an index result of the standard deviation of the strongest intrinsic modal component after the milling signal in (a) is decomposed into a series of intrinsic modal components, and (d) is an index result of the ratio of the chatter-related component of the milling signal in (a) to the standard deviation of the milling signal.

[0048] Figure 5 Example diagram of the chatter state in the present application, wherein (a) is a milling signal of the chatter state, (b) is a spectrum diagram of the milling signal in (a), (c) is an index result of the standard deviation of the strongest intrinsic modal component after the milling signal in (a) is decomposed into a series of intrinsic modal components, and (d) is an index result of the ratio of the chatter-related component of the milling signal in (a) to the standard deviation of the milling signal. DETAILED DESCRIPTION

[0049] The present application will be further described in detail below with reference to the accompanying drawings.

[0050] As shown in the figure, the milling process vibration state monitoring method based on acceleration signals provided by the present application has the following specific steps: Figure 1

[0051] Step 1: Signal acquisition and pretreatment

[0052] Step 1.1: Acquire the acceleration signal during the milling process;

[0053] As shown in the figure, the milling cutter is installed on the spindle of the machine tool, the workpiece is clamped on the workbench, the acceleration sensor is installed in the direction of the normal vector of the surface to be machined of the workpiece, and is connected to the data acquisition device. The acquired acceleration signal is represented by s(n), n represents the signal length, and is transmitted to the computer for vibration state monitoring. Figure 2 Step 1.2: Denoise the acceleration signal s(n) acquired in step 1.1;

[0054]

[0055] ​​Noise estimation is based on the uncut signal segment (signal collected by acceleration sensor when uncut) and the noise in acceleration signal s(n) is eliminated based on this to obtain the pure milling signal. The specific noise reduction process is as follows:

[0056] 1.2.1) The sampled acceleration signal s(n) is the sum of the pure milling signal x(n) and the noise d(n), and the Fourier transform is performed to obtain the expression in the frequency domain:

[0057] s(n) = x(n) + d(n)

[0058] S(ω) = F[s(n)] = X(ω) + D(ω) (1)

[0059] In the formula, F[] represents Fourier transform, ω is the frequency of Fourier component, S(ω), X(ω), and D(ω) represent acceleration signal, milling signal, and noise in the frequency domain respectively.

[0060] 1.2.2) Calculate the amplitude spectrum of the sampled signal S(ω), calculate the average amplitude spectrum of the uncut signal segment using the uncut signal segment, and take the average amplitude spectrum as the amplitude spectrum estimation of the noise signal D(ω);

[0061]

[0062]

[0063] In the formula, |S(ω)| represents the amplitude spectrum of S(ω), φ s (ω) represents the phase of S(ω), represents the amplitude spectrum estimation of D(ω), φ d (ω) represents the phase of D(ω), and j is the imaginary unit.

[0064] 1.2.3) The amplitude spectrum estimation of the pure milling signal in the frequency domain is obtained by amplitude spectrum subtraction Then the amplitude spectrum estimation of the pure milling signal in the frequency domain Inverse Fourier transform is performed to obtain the pure milling signal in the time domain after noise elimination;

[0065]

[0066] with a≥1,0<β<<1

[0067]

[0068] In the formula, represents the amplitude spectrum estimation of X(ω), α represents the over-reduction factor, β represents the spectrum lower limit parameter, and F -1 [] represents inverse Fourier transform.

[0069] Step 2: Vibration state stability judgment

[0070] Step 2.1: Empirical mode decomposition is performed on the pure milling signal x(n) obtained in step 1 to obtain all intrinsic mode components l represents intrinsic mode component IMF i The number of (n) is obtained.

[0071] Empirical mode decomposition is a time-frequency localization analysis method, and empirical mode decomposition is based on time scale characteristics, has time self-adaptability, does not need to pre-set basis function, and is very good at processing nonlinear and non-stationary signals. Therefore, the intrinsic mode components of the milling signal x(n) are obtained by adopting empirical mode decomposition, and the intrinsic mode components obtained after decomposition satisfy the following conditions: 1) the number of cross zero points is equal to the number of extreme points, or the maximum difference is 1, in the whole time range; 2) at any time, the average value of the upper envelope line determined by the local maximum value point and the lower envelope line determined by the local minimum value point is 0.

[0072] The specific decomposition process is as follows:

[0073] 2.1.1 All local maximum value points and local minimum value points of the milling signal x(n) are found, and cubic spline interpolation is used to connect the local maximum value points and the local minimum value points into an upper envelope line e max (n) and a lower envelope line e min (n) respectively;

[0074] 2.1.2 The average values of the upper envelope line e max (n) and the lower envelope line e min (n) at each time are calculated to obtain a mean envelope line

[0075]

[0076] 2.1.3 The first component h1(n) is obtained by subtracting the mean envelope line from the milling signal x(n), and it is judged whether the first component h1(n) satisfies the conditions of the intrinsic mode component, if yes, the first component h1(n) is the first-order intrinsic mode component IMF1(n), if not, step 2.1.4 is performed, if yes, step 2.1.5 is performed;

[0077] 2.1.4 The decomposition component h1(n) is decomposed to obtain the upper envelope line h 1max (n) of h1(n), the lower envelope line h 1min (n) and the mean envelope line of h1(n) are obtained. The component h1(n) is subtracted from the mean envelope line to obtain a component h11 (n) and judge h 11 (n) whether the condition of intrinsic mode component is satisfied, if not, re-decompose the component h1(n) and repeat the step until the component h 1m (n) satisfies the condition of intrinsic mode component, at this time, the component h 1m (n) is the first order intrinsic mode component IMF1(n);

[0078]

[0079] wherein m represents the number of decomposition of the component h1(n);

[0080] 2.1.5 subtract the first order intrinsic mode component IMF1(n) from the milling signal x(n) to obtain the residual r1(n);

[0081] h 1k (n) = IMF1(n)

[0082] r1(n) = x(n) - IMF1(n) (6)

[0083] 2.1.6 take the residual r1(n) as a new original signal, repeat the steps 2.1.1-2.1.5 until the lth intrinsic mode component IMF l (n) corresponds to the residual r l (n) is a monotonic function or a constant, decomposition cannot be further performed, at this time, the milling signal x(n) can be expressed as the sum of the l intrinsic mode components and the residual r l (n), and the decomposition ends.

[0084]

[0085] Step 2.2: according to the energy size, sort the l intrinsic mode components obtained in step 2.1 to find the strongest intrinsic mode component;

[0086] We find that the strongest intrinsic mode component concentrates the most significant vibration state information in the original signal, and this component is most sensitive to the change of the milling signal. Therefore, in order to obtain the strongest intrinsic mode component, it is necessary to sort the intrinsic mode components according to the energy from high to low (or from low to high in other embodiments):

[0087] IMF = sort[IMF1(n), IMF2(n),, IMF l (n)]

[0088]

[0089] wherein sort[] represents energy sorting, representing the energy of the IMF i u1(n), u2(n), u fr (n) represent the frequency bands that make up the IMF i fr is the number of frequency bands.

[0090] Step 2.3: Take the standard deviation of the strongest energy modal component as an identification index of the unstable vibration state to determine the stability of the current vibration state.

[0091] If the vibration is severe, the milling signal amplitude increases, and the amplitude of the strongest energy modal component increases accordingly. Therefore, the standard deviation std of the strongest energy modal component is calculated as an identification index, which reflects the severity of the workpiece vibration. The smaller the value, the more stable the vibration, and the spectrum distribution is uniform. At this time, the workpiece is in a stable vibration state. The larger the value, the more severe the vibration. A fixed threshold L1 is set in advance, L1 is taken in the range of [0.8, 1.2], and the unit is 10g, g is the acceleration of gravity. If the standard deviation std of the strongest energy modal component is less than L1, the current monitoring state is a stable forced vibration state, and then return to step 1 to continue signal acquisition. If the standard deviation std of the strongest energy modal component is greater than or equal to L1, the current monitoring state is an unstable vibration state, and then proceed to step 3.

[0092] wherein the standard deviation of the strongest energy modal component is calculated according to the following formula:

[0093]

[0094]

[0095] In the formula, IMF1(n) represents the amplitude of the strongest energy modal component IMF1 at the nth time, represents the average amplitude of the k-time of the modal component IMF1, and k represents the number of data acquisition points.

[0096] Step 3: Construct a filter to extract the chatter-related component;

[0097] The milling signal x(n) contains the forced vibration component y(n) and the chatter-related component e(n), and a filter needs to be designed to extract the chatter-related component e(n) for identifying the chatter state. Preferably, considering that the signal we collected has the characteristics of non-stationary and non-linear, the least mean square (LMS) filter with adaptive filtering function is designed to extract the chatter-related component e(n) in this embodiment, and the structure is as shown in Figure 3 The design method is as follows:

[0098] Step 3.1: In the milling process, the cutter rotates at a constant speed, and the teeth of the cutter intermittently cut into and out of the material surface, resulting in periodic vibration of the workpiece. Therefore, the vibration frequency of the workpiece is composed of the spindle rotation frequency and its multiples, and the tooth passing frequency and its multiples. The calculation expression of the forced vibration component of the workpiece is as follows:

[0099]

[0100] where y SF (n) is the spindle rotation frequency and its multiples, y TF (n) is the tooth passing frequency and its multiples, c is the multiple, C m is the maximum multiple, is the amplitude, ω SF is the spindle rotation frequency, ω TF is the tooth passing frequency, N s is the spindle speed, z is the number of cutter teeth, Δ T is the sampling time interval, F s is the sampling rate.

[0101] Step 3.2: The input milling signal x(n) is taken as the expected signal, and the forced vibration component y(n) calculated in step 3.1 is taken as the expected signal. Therefore, the composition of the milling signal x(n) is represented as follows:

[0102] x(n) = y(n) + e(n) (11)

[0103] Step 3.3: The estimated value y * (n) of the forced vibration component y(n) and the estimated value e * (n) of the chatter-related component e(n) have the following relationship:

[0104] y * (n) = W(n)x(n)

[0105] e * (n) = y(n) - y * (n) (12)

[0106] where W(n) is the weight vector of the filter;

[0107] Step 3.4: The minimum mean square value of the estimated value e * (n) of the chatter-related component e(n) is taken as the iteration target, and the cost function J(n) is defined. When the cost function J(n) takes the minimum value, the weight coefficients of the filter can reach the optimal value. Therefore, the gradient descent method is used to solve the optimal weight vector, and the iteration formula for the weight vector of the filter is finally obtained as follows:

[0108]

[0109] wherein μ is a step factor, and the value range is (0, 1];

[0110] Step 3.5: Since the forced vibration component y(n) is the main component in the milling signal x(n), when the filter is optimally estimated, the estimated value y * (n) of the forced vibration component y(n) is equal to the forced vibration component y(n), and the estimated value e * (n) of the chatter-related component e(n) is equal to the chatter-related component e(n), at this time, the final weight vector iteration formula can be obtained:

[0111] W(n+1) = W(n) + 2μe * (n)x(n) (14)

[0112] Step 3.6: Through the weight vector iteration formula obtained in step 3.5, the filter is iterated continuously, and finally the forced vibration component y(n) and the chatter-related component e(n) are separated, so that the extraction of the chatter-related component e(n) is realized.

[0113] It should be noted that the present application is not limited to using the least mean square (LMS) filter for adaptive filtering, and in other embodiments, other filters can also be designed, as long as the input of the filter is the main shaft rotation frequency and the gear tooth passing frequency and their multiple frequencies, so as to ensure the accuracy of the extracted chatter component, and then ensure the accuracy of the subsequent chatter identification.

[0114] Step 4: Taking the ratio stdr between the chatter-related component e(n) and the standard deviation of the milling signal x(n) as an identification index of the chatter state, identifying the severe vibration and chatter

[0115] When the chatter is severe, the amplitude of the most sensitive chatter-related component e(n) in the milling signal x(n) increases. The standard deviation of the chatter-related component e(n) and the standard deviation of the milling signal x(n) are calculated respectively, and then the ratio stdr between the two standard deviations is calculated, and the ratio stdr is taken as an identification index of the chatter state. The identification index stdr can reflect the energy of the chatter of the milled workpiece, the smaller the value, the lower the chatter energy, and the larger the value, the higher the chatter energy. A fixed threshold L2 is set in advance, L2 is in the range of [0.7, 1.3], and is dimensionless; when stdr < L2, the current monitoring state is the severe vibration state; when stdr ≥ L2, the current monitoring state is the chatter state.

[0116] The calculation method of the index stdr is as follows:

[0117]

[0118] wherein s0 represents the amplitude standard deviation of the milling signal x(n) within k time, represents the mean value of the amplitude within the previous k time of the milling signal; s1 represents the standard deviation of the amplitude within the previous k time of the chatter-related component e(n), represents the mean value of the amplitude within the previous k time of the chatter-related component e(n).

[0119] Example simulation verification:

[0120] Acceleration signals during milling were collected, and the machining parameters are shown in Table 1. The collected acceleration signals are shown in Figs. 1(a) and 1(b), respectively, where the spindle rotation frequency is SF = 83.3 Hz, and the tooth passing frequency is TF = 333.3 Hz. Figure 4 Figure 5 Acceleration signals during milling were collected, and the machining parameters are shown in Table 1. The collected acceleration signals are shown in Figs. 1(a) and 1(b), respectively, where the spindle rotation frequency is SF = 83.3 Hz, and the tooth passing frequency is TF = 333.3 Hz.

[0121] Table 1 Milling machining parameters

[0122]

[0123] In this example, the identification threshold and principle of the vibration state of the milled workpiece are shown in Table 2. The accuracy of the identification results of the present application is verified by comparing the analysis of the sampled signals using fast Fourier transform (FFT) and the monitoring method of the present application:

[0124] a) In case 1, the frequency spectrum of the sampled signal obtained by fast Fourier transform (FFT) is shown in Fig. 1(b), the spectrum is uniformly distributed, and the peak value is composed of the spindle rotation frequency and the tooth passing frequency. It is determined that this case is in a state of severe vibration. The std and stdr variation curves of this case calculated by the method of the present application are shown in Figs. 1(c) and 1(d), respectively. It can be found that severe vibration occurs throughout the milling process. Figure 4 Figure 4 Figure 4

[0125] b) In case 2, the frequency spectrum of the sampled signal obtained by fast Fourier transform (FFT) is shown in Fig. 1(b), the spectrum is concentrated, and the chatter frequency (CF) appears in the peak value composition. It is determined that this case is in a state of chatter. The std and stdr variation curves of this case calculated by the method of the present application are shown in Figs. 1(c) and 1(d), respectively. It can be found that chatter occurs throughout the milling process. Figure 5 Figure 5 Figure 5

[0126] From the above results, it can be seen that the identification results of the present application are consistent with the results obtained by fast Fourier transform (FFT), proving that the vibration state monitoring method proposed by the present application is feasible, and can accurately identify the unstable vibration state and distinguish whether the vibration state is a severe vibration state or a chatter state.

[0127] Table 2 Identification threshold and principle of the vibration state of the milled workpiece​​​​​​​

[0128]

[0129] The application also provides a milling process vibration state monitoring system based on acceleration signals, which comprises a signal acquisition and preprocessing unit, a vibration state stability determination unit, a chatter component extraction unit and a vibration state recognition unit; the signal acquisition and preprocessing unit is used for acquiring acceleration signals during milling and performing noise reduction processing on the acceleration signals to obtain pure milling signals; the vibration state stability determination unit is used for performing empirical mode decomposition on the milling signals to obtain a series of intrinsic mode components, calculating the standard deviation of the intrinsic mode component with the strongest energy as a judgment index of vibration state stability, comparing the judgment index with a preset threshold value, and determining whether the current vibration state is stable according to the comparison result; the chatter component extraction unit is used for extracting chatter-related components from the milling signals output by the signal acquisition and preprocessing unit and outputting the chatter-related components to the vibration state recognition unit; and the vibration state recognition unit is used for calculating the ratio of the amplitude standard deviation of the chatter-related components to the amplitude standard deviation of the milling signals, comparing the ratio with a preset threshold value as a state recognition index, and recognizing the current unstable state as chatter or severe vibration state according to the comparison result.

[0130] In addition, the application also provides a non-volatile computer readable storage medium, which stores a computer program, and the computer program is used to execute the following steps when being run:

[0131] Step 1: acquiring acceleration signals during milling and performing noise reduction to obtain pure milling signals;

[0132] Step 2: performing empirical mode decomposition on the milling signals to obtain a series of intrinsic mode components, calculating the standard deviation of the intrinsic mode component with the strongest energy, comparing the standard deviation with a preset threshold value, if the standard deviation is less than the threshold value, the current state is a stable forced vibration state, and returning to step 1; if the standard deviation is greater than or equal to the threshold value, the current state is an unstable vibration state, and entering step 3;

[0133] Step 3: extracting chatter-related components from the milling signals;

[0134] Step 4: calculating the ratio of the amplitude standard deviation of the chatter-related components to the amplitude standard deviation of the milling signals, comparing the ratio with a preset threshold value, if the ratio is less than the threshold value, the current state is a severe vibration state; if the ratio is greater than or equal to the threshold value, the current state is a chatter state.

[0135] At the same time, the application also provides an electronic device, which comprises a processor and a storage medium, wherein the storage medium stores a computer program, and the computer program is used to execute the following steps when being run by the processor:

[0136] Step 1: collect the acceleration signal during milling and denoise to obtain a pure milling signal;

[0137] Step 2: empirically mode decomposition of the milling signal to obtain a series of intrinsic mode components, calculate the standard deviation of the intrinsic mode component with the strongest energy, compare the standard deviation with a preset threshold one, if less than the threshold one, the current is a stable forced vibration state, return to step 1; if greater than or equal to the threshold one, the current is an unstable vibration state, enter step 3;

[0138] Step 3: extract the chatter related component from the milling signal;

[0139] Step 4: calculate the ratio of the amplitude standard deviation of the chatter related component to the amplitude standard deviation of the milling signal, compare the ratio with a preset threshold two, if less than the threshold two, the current is a severe vibration state; if greater than or equal to the threshold two, the current is a chatter state.

Claims

1. A method for monitoring the state of vibration of a milling process based on an acceleration signal, characterized in that, The method comprises the following steps: Step 1: collecting the acceleration signal during milling and denoising to obtain a pure milling signal; Step 2: performing empirical mode decomposition on the milling signal to obtain a series of intrinsic mode components, calculating the standard deviation of the intrinsic mode component with the strongest energy, and comparing the standard deviation with a preset threshold one; if the standard deviation is less than the threshold one, the current is a stable forced vibration state, and returning to step 1; if the standard deviation is greater than or equal to the threshold one, the current is an unstable vibration state, and entering step 3; Step 3: separating the milling signal into a forced vibration component and a chatter-related component by using a filter, so as to extract the chatter-related component from the milling signal; Step 4: calculating the ratio of the amplitude standard deviation of the chatter-related component to the amplitude standard deviation of the milling signal, comparing the ratio with a preset threshold two, if the ratio is less than the threshold two, the current is a severe vibration state; if the ratio is greater than or equal to the threshold two, the current is a chatter state.

2. The acceleration signal-based vibration state monitoring method of a milling process according to claim 1, characterized by, The denoising method in step 1 is based on the signal segment without cutting to estimate the noise, and the noise in the acceleration signal is eliminated based on this to obtain a pure milling signal.

3. The acceleration signal-based vibration state monitoring method of a milling process according to claim 2, characterized by, The denoising method in step 1 is specifically: Firstly, the acceleration signal is converted into a frequency domain signal, then the average amplitude spectrum of the frequency domain signal and the amplitude spectrum of the noise signal contained therein are calculated by using the signal segment without cutting, and finally the amplitude spectrum of the milling signal is obtained by operating the average amplitude spectrum and the amplitude spectrum of the noise signal and is inverse transformed to the time domain to obtain a pure milling signal.

4. The method according to any of claims 1 to 3, characterized in that The method for performing empirical mode decomposition on the milling signal in step 2 to obtain a series of intrinsic mode components is specifically: 2.1) Finding the milling signal All local maximum and minimum points are connected by cubic spline interpolation to form the upper envelope and lower envelope respectively 2.2) Calculate upper envelope and lower envelope The average of the envelopes at each time instant gives the mean envelope ; 2.3) with the milling signal Subtract the mean envelope , obtaining the first component , determining whether the first component satisfies the condition of an eigenmode component, if so, the first component is a first-order eigenmode component , if not, step 2.4) is performed, if it is satisfied, step 2.5) is performed; 2.4) decomposition component , the upper envelope , the lower envelope and the mean envelope of , to obtain the component and determine whether the component satisfies the condition of the eigenmode component, if not, repeat the step until the component satisfies the condition of the eigenmode component, at this time the component is the first-order eigenmode component ; 2.5) with a milling signal Subtracting the first order eigenmode component , to obtain a residual ; 2.6) the remainder As a new original signal, steps 2.1) - 2.5) are repeated until the number of eigenmodal components corresponding remainder is a monotonic function or constant, the decomposition ends; The conditions of the intrinsic mode components are: 1) the number of crossing zero points is equal to the number of extreme points, or the maximum difference is 1, in the whole time range; 2) at any time, the average value of the upper envelope line determined by the local maximum points and the lower envelope line determined by the local minimum points is 0.

5. The acceleration signal based milling process vibration condition monitoring method according to claim 4, characterized in that: The filter is a least mean square (LMS) filter with adaptive filtering function.

6. A milling process vibration condition monitoring system based on acceleration signals, characterized by The method comprises: a signal acquisition and preprocessing unit for acquiring the acceleration signal during milling and denoising to obtain a pure milling signal; a vibration state stability determination unit for performing empirical mode decomposition on the milling signal to obtain a series of intrinsic mode components, calculating the standard deviation of the intrinsic mode component with the strongest energy as a vibration state stability evaluation index, and comparing the evaluation index with a preset threshold, and determining whether the current vibration state is stable according to the comparison result; a chatter component extraction unit for separating the milling signal into a forced vibration component and a chatter-related component by using a filter, so as to extract the chatter-related component from the milling signal; a vibration state recognition unit for calculating the ratio of the amplitude standard deviation of the chatter-related component to the amplitude standard deviation of the milling signal, comparing the ratio with a preset threshold as a state recognition index, and recognizing the current unstable state as a chatter or a severe vibration state according to the comparison result.

7. The acceleration signal based milling process vibration condition monitoring system according to claim 6, characterized in that: The signal collection pre-processing unit comprises an acceleration sensor and a data collector connected to the acceleration sensor, the acceleration sensor is installed in the direction of normal vector of the surface of the workpiece to be machined; the chatter component extraction unit comprises a filter, the filter is an adaptive filter.

8. A non-transitory computer readable storage medium having stored thereon a computer program, characterized in that: The computer program is run to execute the milling process vibration state monitoring method of any one of claims 1-5.

9. An electronic device comprising a processor and a storage medium having stored thereon a computer program, characterised in that: The computer program is run by the processor to execute the milling process vibration state monitoring method of any one of claims 1-5.

Images