Micro-milling chatter identification method considering environmental noise effects

Abstract

The present application relates to a kind of micro-milling chatter identification method considering environmental noise influence, the method first uses two accelerometers simultaneously to collect the acceleration signal of workpiece and machine tool side wall position during processing;Then using variable forgetting factor recursive least squares adaptive filter, with the help of machine tool acceleration signal, workpiece acceleration signal is adaptively filtered, remove the environmental noise and periodic component contained in workpiece acceleration signal;Then the power spectral density of filtered workpiece acceleration signal is calculated;Then the corresponding kurtosis, skewness and other characteristics are calculated using the obtained power spectral density, and according to the size of characteristic value, it is compared with the set threshold value, to accurately identify the occurrence of chatter.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of mechanical manufacturing, and relates to a micro-milling chatter identification method, in particular to a micro-milling chatter identification method considering the influence of environmental noise. BACKGROUND

[0002] The existing chatter identification method is for conventional milling, and in conventional milling, the cutting parameters used are large, the vibration signal generated is strong, and the environmental noise is relatively weak, and its influence on the machining process is not considered. For the micro-milling process, the cutting parameters used are small, and the vibration signal generated during machining is weak, at which time the influence of environmental noise cannot be ignored, resulting in that some existing chatter identification methods are no longer applicable. Therefore, it is of important theoretical and practical value to find a chatter identification method suitable for the micro-milling process.

[0003] Document 1“C. Liu, W. Xu, L. Gao, Identification of milling chatter based on a novel frequency domain search algorithm, The International Journal of Advanced Manufacturing Technology 109 (2020) 2393-2407” discloses a conventional milling chatter identification method based on a frequency domain search algorithm. The method considers that when chatter occurs, the difference between the chatter frequency and the tooth frequency harmonic remains constant between adjacent tooth frequency harmonics, and the difference is defined as the vibration fundamental frequency. When the machining process is stable, the above phenomenon does not exist, so the identification of chatter can be realized by judging whether the vibration fundamental frequency exists. However, the above method is for conventional milling, and the cutting parameters used in the experiment are large, the vibration signal generated during machining is strong, and the environmental noise is relatively weak, so the influence of environmental noise is not considered, and for the micro-milling process, the existence of environmental noise will lead to the misjudgment of the vibration fundamental frequency.

[0004] Document 2 "Y. Chen, H. Li, X. Jing, L. Hou, X. Bu, Intelligent chatter detection using image features and support vector machine, The International Journal of Advanced Manufacturing Technology 102 (2019) 1433-1442" discloses a conventional milling chatter recognition method based on fast Fourier transform and short-time Fourier transform. The method first performs additional experiments, then uses fast Fourier transform to analyze them in order to determine the chatter dominant frequency band, and based on the chatter dominant frequency band, the time-frequency image obtained by short-time Fourier transform is segmented to obtain the corresponding sub-image, and then the mean, variance and other features corresponding to the sub-image grayscale image are calculated to realize the recognition of chatter. However, the above method considers that in conventional milling, the influence of environmental noise is not obvious, so it does not consider environmental noise, and for micro-milling process, the existence of environmental noise will lead to false chatter dominant frequency band.

[0005] The typical feature of the above document is that the proposed method is for conventional milling and does not consider the influence of environmental noise. In the micro-milling process, the cutting parameters used are small, and the amplitude of the vibration signal generated during the machining process is close to the amplitude of the environmental noise, so the environmental noise will have a non-negligible influence on the micro-milling process and seriously interfere with the recognition of micro-milling chatter. SUMMARY

[0006] Technical problems to be solved

[0007] In view of the defects that the existing chatter recognition method does not consider the influence of environmental noise when the cutting parameter is small, which leads to its inability to be directly applied to the micro-milling process, the present application proposes a micro-milling chatter recognition method suitable for micro-milling process and considering the influence of environmental noise.

[0008] Technical scheme

[0009] A micro-milling chatter recognition method considering the influence of environmental noise, characterized in that the steps are as follows:

[0010] Step 1: Perform experiments with set cutting parameters, and simultaneously collect workpiece acceleration signals and machine tool acceleration signals during the experiments, wherein the workpiece acceleration signal corresponding to the machining stage is d(n), and the machine tool acceleration signal is m(n), wherein n represents the time index;

[0011] Step 2: The workpiece acceleration signal d(n) and the machine acceleration signal m(n) are sent into a variable-forgetting-factor recursive least square adaptive filter to obtain the filtered workpiece acceleration signal r(n):

[0012]

[0013] where, is the weight vector, x(n) = [m(n), m(n-1),..., m(n-L+1)] T is the input signal vector, and L is the filter order;

[0014] Step 3: Update the weight vector

[0015]

[0016]

[0017] where k(n) is the Kalman gain vector, λ(n) is the forgetting factor, and P(n-1) is the autocorrelation matrix of the input signal vector x(n-1);

[0018] Step 4: Update the forgetting factor λ(n) according to the following formula:

[0019]

[0020]

[0021]

[0022]

[0023] q(n) = x T (n)P -1 (n-1)x(n)

[0024] where q(n) is an intermediate variable, is the power of q(n), is a transition variable, is the power of the filtered workpiece acceleration signal r(n), δ is a very small positive number to prevent the denominator from being equal to zero, a p is the axial cutting depth used in the experiment, a p_max is the maximum value of the axial cutting depth used in the experiment, ξ(n) is the convergence factor, and ε is the set convergence threshold, is the power of the workpiece acceleration signal, λ max is the maximum value of the forgetting factor, is the output signal of the filter, is the output signal of the filter the power of the filtered workpiece acceleration signal r(n) is calculated using the following formula:

[0025] the power of the filtered workpiece acceleration signal r(n) is calculated using the following formula:

[0026]

[0027] where Δt is the sampling interval, N is the number of data points, e is the natural constant, j is the imaginary unit, f is the frequency, S i is the spectral energy corresponding to the frequency f;

[0028] the kurtosis value f1 and the skewness value f2 corresponding to the power spectral density are calculated using the following formula:

[0029]

[0030]

[0031] Step 7: compare the result obtained in step 6 with the set threshold value, if higher than the threshold value, it means that chatter occurs in the machining process; if lower than the threshold value, it means that the machining process is stable.

[0032] A computer system, characterized by comprising: one or more processors, a computer readable storage medium for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the above method.

[0033] A computer readable storage medium, characterized by storing computer executable instructions, the instructions when executed are used to implement the above method.

[0034] Advantages

[0035] The micro-milling chatter identification method considering the influence of environmental noise provided by the application first obtains the acceleration signals at the workpiece and the machine tool side wall positions during the machining process through two accelerometers; then uses a variable forgetting factor recursive least squares adaptive filter to filter out the environmental noise and periodic components contained in the workpiece acceleration signal; further calculates the power spectral density corresponding to the filtered workpiece acceleration signal, and based on the obtained power spectral density, calculates the corresponding kurtosis, skewness and other characteristics; compare it with the set threshold value, the identification of micro-milling chatter can be realized.

[0036] Compared with document 1, the application considers the influence of environmental noise on the micro-milling machining process, can effectively filter out the environmental noise, and obtain more accurate chatter identification results; compared with document 2, the environmental noise and periodic components and other interference signals are filtered out when the chatter identification is performed, so it is not necessary to determine the chatter dominant frequency band in advance. BRIEF DESCRIPTION OF DRAWINGS

[0037] The accompanying drawings are included to provide a further understanding of the application and are incorporated in and constitute a part of this specification. The drawings illustrate embodiments of the application and, together with the description, serve to explain the principles of the application. In the drawings:

[0038] Figure 1 is the workpiece acceleration signal obtained when the spindle speed of the embodiment of the application is 18000 rpm and the axial cutting depth is 10 microns.

[0039] Figure 2 is the machine acceleration signal obtained when the spindle speed of the embodiment of the application is 18000 rpm and the axial cutting depth is 10 microns.

[0040] Figure 3 is the filtered workpiece acceleration signal obtained when the spindle speed of the embodiment of the application is 18000 rpm and the axial cutting depth is 10 microns using a variable-forgetting-factor recursive least squares adaptive filter.

[0041] Figure 4 is the power spectral density corresponding to the filtered workpiece acceleration signal when the spindle speed of the embodiment of the application is 18000 rpm and the axial cutting depth is 10 microns.

[0042] Figure 5 is the workpiece acceleration signal obtained when the spindle speed of the embodiment of the application is 18000 rpm and the axial cutting depth is 200 microns.

[0043] Figure 6 is the machine acceleration signal obtained when the spindle speed of the embodiment of the application is 18000 rpm and the axial cutting depth is 200 microns.

[0044] Figure 7 is the filtered workpiece acceleration signal obtained when the spindle speed of the embodiment of the application is 18000 rpm and the axial cutting depth is 200 microns using a variable-forgetting-factor recursive least squares adaptive filter.

[0045] Figure 8 is the power spectral density corresponding to the filtered workpiece acceleration signal when the spindle speed of the embodiment of the application is 18000 rpm and the axial cutting depth is 200 microns. DETAILED DESCRIPTION

[0046] In order to make the objects, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and should not be used to limit the present application. In addition, the technical features involved in each embodiment of the application described below can be combined with each other as long as they do not conflict with each other.

[0047] The micro-milling chatter identification method considering the influence of environmental noise provided by the application firstly collects the acceleration signals of the workpiece and the position of the machine tool side wall during the machining process by using two accelerometers; then the workpiece acceleration signal is adaptively filtered by using the machine tool acceleration signal by means of a variable forgetting factor recursive least square adaptive filter, so as to remove the environmental noise and periodic components contained in the workpiece acceleration signal; then the power spectral density of the filtered workpiece acceleration signal is calculated; then the corresponding kurtosis, skewness and other characteristics are calculated by using the obtained power spectral density, and the characteristic values are compared with the set threshold value according to the size, so as to accurately identify the occurrence of chatter.

[0048] The technical effect expected by the application: The micro-milling chatter identification method provided by the application considers the influence of environmental noise when the cutting parameter is small, and can accurately identify the occurrence of micro-milling chatter.

[0049] Example 1 corresponds to a stable micro-milling machining process, and example 2 corresponds to a micro-milling machining process with chatter.

[0050] Example 1:

[0051] In the experiment, a flat-bottomed carbide end mill with tooth number N=2 teeth and radius R=0.5 millimeters is used for aluminum alloy micro-milling test on a precision engraving GR50 cradle five-axis numerical control machining center in the down-milling cutting mode. The cutting parameters used are radial cutting depth a e =0.5 millimeters, feed per tooth c=0.005 millimeters / tooth, spindle speed 18000 revolutions / minute, and axial cutting depth a p =10 microns.

[0052] Step one, refer to the attached Figures 1-2 The experiment is carried out with the set cutting parameters, and the workpiece acceleration signal and the machine tool acceleration signal are collected simultaneously during the experiment, wherein the workpiece acceleration signal corresponding to the machining stage is d(n), and the machine tool acceleration signal is m(n), wherein n=1, 2, …, 800000.

[0053] Step two, the workpiece acceleration signal d(n) and the machine tool acceleration signal m(n) are sent into a variable forgetting factor recursive least square adaptive filter to obtain the filtered workpiece acceleration signal r(n), refer to the attached Figure 3 :

[0054]

[0055] In the formula, is the weight vector, x(n)=[m(n), m(n-1), …, m(n-L+1)] T is the input signal vector, and the filter order L=5.

[0056] Step 3: Update the weight vector

[0057]

[0058]

[0059] Where k(n) is the Kalman gain vector, λ(n) is the forgetting factor, and P(n-1) is the autocorrelation matrix of the input signal vector x(n-1). After the filter converges, the final weight vector is obtained.

[0060] Step 4: Update the forgetting factor λ(n) according to the following formula:

[0061]

[0062]

[0063]

[0064]

[0065] q(n)=x T (n)P -1 (n-1)x(n)

[0066] In the formula, q(n) is an intermediate variable. It is the power of q(n). It is a transitional variable. This is the power of the filtered workpiece acceleration signal r(n), δ=0.00000001 to prevent the denominator from being equal to zero. The axial cutting depth a used in the example. p =10 micrometers, the maximum value of the axial cutting depth used in the experiment, a p_max =300 micrometers, ξ(n) is the convergence factor, and the set convergence threshold ε = 0.00001. It is the power of the workpiece acceleration signal, and the maximum value of the forgetting factor λ. max =0.9, It is the output signal of the filter. It is the output signal of the filter. The power.

[0067] Step 5: Calculate the power spectral density of the filtered workpiece acceleration signal r(n) using the following formula, refer to the appendix. Figure 4 :

[0068]

[0069] where sampling interval Δt = 1 / 51200, data point number N = 800000, natural constant e = 2.718, j is imaginary unit, f is frequency, S i is the spectral energy corresponding to frequency f.

[0070] Step six, calculate the kurtosis value f1 and skewness value f2 corresponding to the power spectral density by using the following formula:

[0071]

[0072]

[0073] The kurtosis value f1 = 23.007 and the skewness value f2 = 4.690 are calculated.

[0074] Step seven, compare the results obtained in step six with the set threshold value, here, the threshold values of kurtosis and skewness are set to 45 and 22 respectively, and the comparison shows that both the kurtosis value and the skewness value are less than the corresponding threshold values, therefore, the machining process is stable.

[0075] Example 2:

[0076] The experiment selects a flat-bottomed carbide end mill with tooth number N = 2 teeth and radius R = 0.5 mm to conduct aluminum alloy micro-milling test on a GR50 cradle five-axis CNC machining center using the face milling cutting method. The cutting parameters used are radial depth of cut a e = 0.5 mm, feed per tooth c = 0.005 mm / tooth, spindle speed 18000 rpm, axial cutting depth a p = 200 microns.

[0077] Step one, refer to the attached Figures 5-6 , the experiment is carried out with the set cutting parameters, and the workpiece acceleration signal and machine tool acceleration signal are collected simultaneously during the experiment, where the workpiece acceleration signal corresponding to the machining stage is d(n), and the machine tool acceleration signal is m(n), where n = 1, 2, …, 800000.

[0078] Step two, send the workpiece acceleration signal d(n) and the machine tool acceleration signal m(n) into the variable forgetting factor recursive least square adaptive filter to obtain the filtered workpiece acceleration signal r(n), refer to the attached Figure 7 :

[0079]

[0080] where, is the weight vector, x(n) = [m(n), m(n-1), …, m(n-L+1)] T is the input signal vector, and the filter order L = 5.

[0081] Step three, update the weight vector

[0082]

[0083]

[0084] where k(n) is the Kalman gain vector, λ(n) is the forgetting factor, and P(n-1) is the autocorrelation matrix of the input signal vector x(n-1). After the filter converges, the final weight vector is obtained as

[0085] Step four, update the forgetting factor λ(n) according to the following equation:

[0086]

[0087]

[0088]

[0089]

[0090] q(n) = x T (n) P -1 (n-1) x(n)

[0091] where q(n) is an intermediate variable, is the power of q(n), is a transition variable, is the power of the filtered workpiece acceleration signal r(n), δ = 0.00000001 to prevent the denominator from being equal to zero, the axial cutting depth a used in the example p = 200 microns, the maximum axial cutting depth a used in the experiment p_max = 300 microns, ξ(n) is the convergence factor, the convergence threshold ε = 0.00001 is set, is the power of the workpiece acceleration signal, the maximum forgetting factor λ max = 0.9, is the output signal of the filter, is the power of the output signal of the filter .

[0092] Step five, calculate the power spectral density of the filtered workpiece acceleration signal r(n) using the following equation, refer to the attached Figure 8 :

[0093]

[0094] In the formula, sampling interval Δt=1 / 51200, data point number N=800000, natural constant e=2.718, j is imaginary unit, f is frequency, S i is the spectral energy corresponding to the frequency f.

[0095] Step six, the kurtosis value f1 and the skewness value f2 corresponding to the power spectral density are calculated by using the following formula:

[0096]

[0097]

[0098] The kurtosis value f1=1832.429 and the skewness value f2=30.067 are calculated.

[0099] Step seven, the result obtained in step six is compared with the set threshold value, here, the threshold values of the kurtosis value and the skewness value are set to 45 and 22 respectively, and the comparison shows that the kurtosis value and the skewness value are both greater than the corresponding threshold values, which indicates that the chatter occurs in the machining process.

[0100] As can be seen from the above example, the proposed method can effectively filter out the environmental noise and periodic components contained in the workpiece acceleration signal, the kurtosis value and the skewness value of the power spectral density of the filtered workpiece acceleration signal are calculated, and they are compared with the corresponding threshold values, thereby realizing the accurate identification of the micro-milling chatter.

[0101] The above is only a specific embodiment of the present application, but the protection scope of the present application is not limited to this, any skilled person in the art can easily think of various equivalent modifications or replacements within the technical range disclosed in the present application, and these modifications or replacements should be covered in the protection scope of the present application.

Claims

1. A method for identifying chatter in micromilling considering the influence of environmental noise, characterized in that... The steps are as follows: Step 1: Conduct the experiment with the set cutting parameters, and simultaneously collect the workpiece acceleration signal and the machine tool acceleration signal during the experiment. The workpiece acceleration signal corresponding to the machining stage is d(n), and the machine tool acceleration signal is m(n), where n represents the time index. Step 2: Input the workpiece acceleration signal d(n) and the machine tool acceleration signal m(n) into a recursive least squares adaptive filter with a variable forgetting factor to obtain the filtered workpiece acceleration signal r(n): In the formula, It is a weight vector, x(n) = [m(n), m(n-1), ..., m(n-L+1)] T L is the input signal vector, and L is the filter order. Step 3: Update the weight vector Where k(n) is the Kalman gain vector, λ(n) is the forgetting factor, and P(n-1) is the autocorrelation matrix of the input signal vector x(n-1); Step 4: Update the forgetting factor λ(n) according to the following formula: q(n)=x T (n)P -1 (n-1)x(n) In the formula, q(n) is an intermediate variable. It is the power of q(n). It is a transitional variable. This is the power of the filtered workpiece acceleration signal r(n), where δ is a very small positive number to prevent the denominator from equaling zero, and a p This refers to the axial cutting depth used in the experiment, a. p_max ξ(n) is the maximum value of the axial cutting depth used in the experiment, ξ(n) is the convergence factor, and ε is the set convergence threshold. It is the power of the workpiece acceleration signal, λ max It is the maximum value of the forgetting factor. It is the output signal of the filter. It is the output signal of the filter. The power; Step 5: Calculate the power spectral density of the filtered workpiece acceleration signal r(n) using the following formula: In the formula, Δt is the sampling interval, N is the number of data points, e is the natural constant, j is the imaginary unit, f is the frequency, and S... i It is the spectral energy corresponding to frequency f; Step 6: Calculate the kurtosis value f1 and skewness value f2 corresponding to the power spectral density using the following formula: Step 7: Compare the result obtained in Step 6 with the set threshold. If it is higher than the threshold, it indicates that chattering has occurred during the processing; if it is lower than the threshold, it indicates that the processing is stable.

2. A computer system, characterized in that... include: One or more processors, a computer-readable storage medium for storing one or more programs, wherein, when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the method of claim 1.

3. A computer-readable storage medium, characterized in that... The device stores computer-executable instructions, which, when executed, are used to implement the method of claim 1.

Images