Milling chatter feature extraction method based on synchronous hierarchical entropy
Through the milling flutter feature extraction method based on synchronous hierarchical entropy, the shortcomings of the prior art in processing nonlinear and non-stationary signals are solved, and more accurate and reliable flutter feature extraction is achieved, which improves the monitoring and control effect of the milling process.
Patent Information
- Application Number
- CN202510521175.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-24
- Publication Date
- 2025-08-08
AI Technical Summary
Existing milling flutter feature extraction methods such as Fourier transform and wavelet transform are poor in processing nonlinear and non-stationary signals, and commonly used entropy methods are difficult to characterize various flutter types, resulting in insufficient accuracy of milling flutter analysis and recognition.
The milling flutter feature extraction method based on synchronous hierarchical entropy is adopted. By selecting the sensor type and installation location, synchronous vibration data are obtained, and multi-layer decomposition is performed to calculate the synchronous hierarchical entropy of each layer node as the flutter feature.
It can reliably extract multiple milling flutter features, adapt to different working conditions, has good anti-noise interference ability, and improves the accuracy and applicability of flutter recognition.
Smart Images
Figure CN120448778A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of mechanical system fault feature extraction, and in particular relates to a milling chatter feature extraction method based on Synchronous Hierarchical Entropy (SHEn). Background Art
[0002] Milling chatter is a self-excited vibration phenomenon caused by the interaction between the tool and the workpiece, typically exhibiting nonlinear and nonstationary dynamic behavior during the cutting process. The occurrence of milling chatter is typically closely related to cutting forces, workpiece rigidity, tool vibration characteristics, and cutting parameters such as feed rate and depth of cut. When the tool contacts the workpiece, the cutting forces continuously change, potentially triggering chatter. Milling chatter not only negatively impacts machining quality, accuracy, and tool life, but can also lead to equipment damage and reduced production efficiency. Therefore, milling chatter analysis and identification are necessary to avoid its harmful effects.
[0003] When it comes to milling chatter analysis and identification, chatter feature extraction is one of the most important tasks. Currently, the most commonly used method for milling chatter feature extraction is the Fourier transform. By converting the signal from the time domain to the frequency domain, the frequency characteristics of the signal can be extracted, which can then be used to analyze and identify chatter. However, the Fourier transform is a linear signal processing method that is relatively weak in processing nonlinear and nonstationary signals and lacks the ability to capture signal frequency variations.
[0004] Another commonly used method for milling chatter feature extraction is the wavelet transform. The wavelet transform effectively processes signals in both the time and frequency domains and is suitable for handling nonlinear and nonstationary signals. While the wavelet transform excels in milling chatter feature extraction, its performance still depends on the selected wavelet basis function. An inappropriate wavelet basis function can lead to erroneous results.
[0005] In recent years, with the continuous development of entropy methods, entropy-based methods for milling chatter feature extraction have attracted considerable attention. Common entropy methods include sample entropy, fuzzy entropy, permutation entropy, and diversity entropy. However, these entropy methods cannot effectively characterize various chatter types.
[0006] To overcome the shortcomings of the aforementioned entropy methods in chatter feature extraction, the present invention discloses a novel synchronous hierarchical entropy (SHEn) and proposes a milling chatter feature extraction method based on this SHEn. The milling chatter feature extraction method based on SHEn, disclosed in this patent, takes into account the chatter frequency characteristics and can extract more accurate and reliable dynamic features from nonlinear and non-stationary vibration signals, providing a new technical means for real-time monitoring and intelligent control during the milling process. With further research and application, the method of the present invention is expected to play an increasingly important role in modern manufacturing, improving machining accuracy, optimizing machining processes, and extending tool life. Summary of the Invention
[0007] Purpose of the Invention: This patent discloses a method for extracting milling chatter features based on synchronous hierarchical entropy. Using synchronous hierarchical entropy as a fault signature, the combing filtering properties of synchronous hierarchical entropy allow it to extract chatter information both synchronous and asynchronous with the rotational speed, thereby effectively characterizing chatter faults. This method, using synchronous hierarchical entropy, enables real-time and reliable chatter feature extraction.
[0008] Technical solution: This invention patent discloses a milling chatter feature extraction method based on synchronous hierarchical entropy, which specifically includes the following steps:
[0009] Step (1): Select the sensor type and installation location, and set relevant parameters;
[0010] Step (2): Obtaining synchronous vibration data X;
[0011] Step (3): Determine the decomposition operator and node X of the n-layer n,e ;
[0012] Step (4): Decompose the synchronous vibration data X into n layers to obtain the low-frequency component and high-frequency component. and regard each component as a node;
[0013] Step (5): Calculate the synchronous hierarchical entropy of each node obtained from each layer of decomposition;
[0014] Step (6): The synchronization level entropy calculated in step (5) is regarded as the chatter feature.
[0015] Furthermore, the implementation process of step (1) is as follows:
[0016] The sensors are vibration sensors and speed sensors; the vibration sensor is installed on the workpiece or the spindle to measure the vibration of the workpiece or the spindle; the speed sensor is installed near the spindle or the milling cutter to measure the speed of the spindle or the milling cutter; the milling cutter is installed on the spindle, and the milling cutter speed is the same as the spindle speed; the relevant parameters include: the number of milling cutter teeth L, the number of synchronous sampling points M in each tooth cycle of the milling cutter, the sampling length, the number of intervals ε, and the number of decomposition layers n; the sampling length is kLM, where k is the number of spindle revolutions.
[0017] Furthermore, the implementation process of step (2) is as follows:
[0018] The spindle speed is used as a reference signal, and a vibration sensor is used to obtain synchronous vibration data during the milling process. The obtained synchronous vibration data is synchronous vibration displacement data, synchronous vibration velocity data, or synchronous vibration acceleration data. The obtained synchronous vibration data can be expressed as:
[0019] X=[x(1),x(2),…,x(i),…,x(kLM)].
[0020] Furthermore, the implementation process of step (3) is as follows:
[0021] Determine the decomposition operator Q j :
[0022]
[0023] Among them, when j = 0 and 1, the low-frequency decomposition operator Q0 and the high-frequency decomposition operator Q1 are obtained respectively, as follows:
[0024]
[0025] Wherein, i=1, 2, ..., (kL-n)M;
[0026] The determination node X n,e The implementation process is as follows: construct a vector [m1,m2,…,m j ,…,m n ], where m j =0 or 1, use [m1,m2,…,m j ,…,m n ] to express an integer e, where e has the following form: n is the number of decomposition levels, e is a non-negative integer, and for a given e there is a unique vector [m1,m2,…,m j ,…,m n ] corresponds to it;
[0027] According to the decomposition operator Q j Define the decomposition operator at the nth layer as follows:
[0028]
[0029] Among them, each row element of the matrix to element The interval is M; when j = 0, the low-frequency decomposition operator is obtained, and when j = 1, the high-frequency decomposition operator is obtained.
[0030] Furthermore, the implementation process of step (4) is as follows:
[0031] Using decomposition operator Decompose the synchronous vibration data X to the nth level, and take the number of synchronous sampling points M in each tooth cycle of the milling cutter as the interval to obtain the low-frequency component and high-frequency component of the nth level, which is 2 n Each component is considered a node and marked as X n,e ,
[0032]
[0033] The n-layer decomposition yields decomposition components.
[0034] Furthermore, the implementation process of step (5) is as follows:
[0035] Calculate the synchronous hierarchical entropy of each node obtained by each layer decomposition; for the nth layer node X n,e Synchronous hierarchical entropy SHEn n,e The calculation process is as follows: first, the time delay is set to the time to obtain a data, the embedding dimension is set to 2, and the n-th layer node X is reconstructed. n,e The phase space of n,e ; Then calculate the phase space X′ n,e Two adjacent vectors v j and v j+1 Cosine similarity of as follows:
[0036]
[0037] A j =v j ·v j+1 ,
[0038] B j =||v j ||,
[0039] v j =[x n,e (j),x n,e (j+1)],
[0040] where j = 1, 2, 3, ..., (kL-n)M-1, v j is the jth vector in phase space, x n,e (j) is the nth layer node X n,e The jth element of ;
[0041] Then divide the interval [-1,1] into ε intervals (I1,I2,…I ε ), calculate the above cosine similarity respectively The probability of being in each interval in, According to the probability of each interval Calculate the synchronous hierarchical entropy SHEn n,e , which is calculated as follows:
[0042]
[0043] Finally, for the n-layer decomposition decomposition components, we can get Synchronous level entropy.
[0044] Furthermore, the implementation process of step (6) is as follows:
[0045] Calculate the result from step (5) The synchronous level entropy is regarded as the chatter feature, thus obtaining A vibration characteristic.
[0046] Beneficial effects: Compared with the prior art, the present invention mainly includes the following beneficial effects: (1) The milling chatter feature extraction method based on synchronous hierarchical entropy disclosed in the present invention can reliably extract multiple milling chatter features; (2) The milling chatter feature extraction method based on synchronous hierarchical entropy disclosed in the present invention can be applied to milling processing under different working conditions, and has good adaptability and versatility; (3) The milling chatter feature extraction method based on synchronous hierarchical entropy disclosed in the present invention has strong anti-noise interference ability, can extract chatter features under strong noise background, and improve the accuracy of chatter recognition. BRIEF DESCRIPTION OF THE DRAWINGS
[0047] Figure 1 This is a flow chart of the milling chatter feature extraction method based on synchronous hierarchical entropy disclosed by the present invention;
[0048] Figure 2 This is the synchronous vibration data when there is no chatter;
[0049] Figure 3 Synchronous vibration data for period-2 bifurcation type chatter;
[0050] Figure 4 This is the synchronous vibration data of the period-3 bifurcation type flutter;
[0051] Figure 5 Synchronous vibration data for Hopf bifurcation type flutter;
[0052] Figure 6 It is the extraction result of the two synchronous level entropies obtained by the first level decomposition;
[0053] Figure 7 The extraction results of the three synchronous level entropies obtained by the second level decomposition; DETAILED DESCRIPTION
[0054] The present invention will be described in further detail below with reference to the accompanying drawings.
[0055] The calculation process of the milling chatter feature extraction method based on synchronous hierarchical entropy disclosed in the present invention is as follows: Figure 1 As shown, the specific steps include:
[0056] Step 1: Select the sensor type and installation location, and set the relevant parameters.
[0057] The sensors are vibration sensors and speed sensors; the vibration sensor is installed on the workpiece or the spindle to measure the vibration of the workpiece or the spindle; the speed sensor is installed near the spindle or the milling cutter to measure the speed of the spindle or the milling cutter; the relevant parameters include: the number of milling cutter teeth L, the number of synchronous sampling points M per tooth cycle of the milling cutter, the sampling length, the number of intervals ε, and the number of decomposition layers n; the sampling length is kLM, where k is the number of spindle revolutions.
[0058] Step 2: Obtain synchronized vibration data X.
[0059] The spindle speed is used as a reference signal, and a vibration sensor is used to obtain synchronous vibration data during the milling process. The obtained synchronous vibration data is synchronous vibration displacement data, synchronous vibration velocity data, or synchronous vibration acceleration data. The obtained synchronous vibration data can be expressed as:
[0060] X=[x(1),x(2),…,x(i),…,x(kLM)].
[0061] Figure 2-5 The synchronous vibration data of no flutter, period-2 bifurcation type flutter, period-3 bifurcation type flutter, and Hopf bifurcation type flutter are displayed respectively.
[0062] Step 3: Determine the decomposition operator and node X of the n-th layer n,e .
[0063] Determine the decomposition operator Q j :
[0064]
[0065] Among them, when j = 0 and 1, the low-frequency decomposition operator Q0 and the high-frequency decomposition operator Q1 are obtained respectively, as follows:
[0066]
[0067] Wherein, i=1, 2, ..., (kL-n)M;
[0068] The determination node X n,e The implementation process is as follows: construct a vector [m1,m2,…,m j ,…,m n ], where m j =0 or 1, use [m1,m2,…,m j ,…,m n ] to express an integer e, where e has the following form: n is the number of decomposition levels, e is a non-negative integer, and for a given e there is a unique vector [m1,m2,…,m j ,…,m n ] corresponds to it;
[0069] According to the decomposition operator Q j Define the decomposition operator at the nth layer as follows:
[0070]
[0071] Among them, the elements of each row of the matrix to element The interval is M; when j = 0, the low-frequency decomposition operator is obtained, and when j = 1, the high-frequency decomposition operator is obtained.
[0072] Step 4: Decompose the synchronous vibration data X into n layers to obtain the low-frequency component and high-frequency component. and regard each component as a node.
[0073] Using decomposition operator The synchronous vibration data X is decomposed at the nth level, and the number of synchronous sampling points M in each tooth cycle of the milling cutter is used as the interval, and the low-frequency component and high-frequency component obtained by the nth level decomposition are 2 n Each component is considered a node and marked as X n,e ,
[0074]
[0075] The n-layer decomposition yields decomposition components.
[0076] The low-frequency and high-frequency components obtained by decomposition in the first layer are as follows:
[0077]
[0078] Until the pre-set decomposition level number n is completed, where M is the number of synchronous sampling points in each tooth of the milling cutter, k is the number of spindle revolutions, and L is the number of milling cutter teeth.
[0079] Step 5: Calculate the synchronous hierarchical entropy of each node obtained from each layer of decomposition.
[0080] Calculate the synchronous hierarchical entropy of each node obtained by each layer decomposition; for the nth layer node X n,e Synchronous hierarchical entropy SHEn n,e The calculation process is as follows: first, the time delay is set to the time to obtain a data, the embedding dimension is set to 2, and the n-th layer node X is reconstructed. n,e The phase space of n,e ; Then calculate the phase space X′ n,e Two adjacent vectors v j and v j+1 Cosine similarity of as follows:
[0081]
[0082] A j =v j ·v j+1 ,
[0083] B j =||v j ||,
[0084] v j =[x n,e (j),x n,e (j+1)],
[0085] where j = 1, 2, 3, ..., (kL-n)M-1, v j is the jth vector in phase space, x n,e (j) is the nth layer node X n,e The jth element of ;
[0086] Then divide the interval [-1,1] into ε intervals (I1,I2,…I ε ), calculate the above cosine similarity respectively The probability of being in each interval in, According to the probability of each interval Calculate the synchronous hierarchical entropy SHEn n,e , which is calculated as follows:
[0087]
[0088] Finally, for the n-layer decomposition decomposition components, we can get Synchronous level entropy.
[0089] Step 6: The synchronization level entropy calculated in step (5) is regarded as a chatter indicator.
[0090] Calculate the result from step (5) The synchronous level entropy is regarded as the chatter feature, thus obtaining A vibration characteristic.
[0091] Figure 6 The results of extracting two synchronous level entropies from the first-level decomposition are shown. As can be seen from the figure, these two synchronous level entropies can distinguish between the no-chatter state and the three chatter states. Figure 7 The results of extracting the three synchronous level entropies obtained by the second level decomposition are shown. In the figure, the coordinate axes x, y, and z correspond to the three synchronous level entropies SHEn obtained by the second level decomposition. 2,0 ,SHEn 2,1 and SHen 2,3 As can be seen from the figure, the three extracted synchronous level entropies can completely distinguish the chatter-free state and the three chatter states.
[0092] The synchronous hierarchical entropy extracted by the method of the present invention can effectively reflect the dynamic change characteristics of the milling processing signal at different levels, thereby providing an important basis for vibration detection, diagnosis and control.
[0093] In this embodiment, a hierarchical decomposition method is used to process synchronously sampled data. High-frequency and low-frequency decomposition of the synchronous vibration data is performed using the number of synchronous sampling points M per tooth, rather than directly performing hierarchical decomposition on two adjacent data points. This method leverages the comb filtering properties of the hierarchical decomposition method to effectively extract milling chatter characteristics, overcoming the limitation of traditional hierarchical entropy methods in characterizing milling chatter types. The chatter characteristics extracted by this method can effectively characterize milling chatter.
Claims
1. Milling chatter feature extraction method based on synchronous hierarchical entropy, characterized by: The following steps are involved: (1): Select the sensor type and installation location, and set the relevant parameters; (2): Obtain synchronous vibration data X; (3): Determine the decomposition operator and node X of the n-layer n,e ; (4): Decompose the synchronous vibration data X into n layers, and obtain the low-frequency component and high-frequency component. and regard each component as a node; (5): Calculate the synchronous hierarchical entropy of each node obtained from each layer of decomposition; (6): The synchronization level entropy calculated in step (5) is regarded as the chatter feature.
2. The milling chatter feature extraction method based on synchronous hierarchical entropy according to claim 1 is characterized in that: The implementation process of step (1) is as follows: The sensors are vibration sensors and speed sensors; the vibration sensor is installed on the workpiece or the spindle to measure the vibration of the workpiece or the spindle; the speed sensor is installed near the spindle or the milling cutter to measure the speed of the spindle or the milling cutter; the milling cutter is installed on the spindle, and the milling cutter speed is the same as the spindle speed; the relevant parameters include: the number of milling cutter teeth L, the number of synchronous sampling points M in each tooth cycle of the milling cutter, the sampling length, the number of intervals ε, and the number of decomposition layers n; the sampling length is kLM, where k is the number of spindle revolutions.
3. The milling chatter feature extraction method based on synchronous hierarchical entropy according to claim 1 is characterized in that: The implementation process of step (2) is as follows: The spindle speed is used as a reference signal, and a vibration sensor is used to obtain synchronous vibration data during the milling process. The obtained synchronous vibration data is synchronous vibration displacement data, synchronous vibration velocity data, or synchronous vibration acceleration data. The obtained synchronous vibration data can be expressed as: X=[x(1),x(2),…,x(i),…,x(kLM)].
4. The milling chatter feature extraction method based on synchronous hierarchical entropy according to claim 1 is characterized in that: In step (3), the decomposition operator Q is determined j as follows: Among them, when j = 0 and 1, the low-frequency decomposition operator Q0 and the high-frequency decomposition operator Q1 are obtained respectively, as follows: Wherein, i=1, 2, ..., (kL-n)M; The determination node X n,e The implementation process is as follows: construct vector where m j =0 or 1, use to express an integer e, where e has the following form: n is the number of decomposition levels, e is a non-negative integer, and for a given e there is a unique vector Corresponding to this; According to the decomposition operator Q j Define the decomposition operator at the nth layer as follows: Among them, each row element of the matrix to element The interval is M; when j = 0, the low-frequency decomposition operator is obtained, and when j = 1, the high-frequency decomposition operator is obtained.
5. The milling chatter detection method based on synchronous hierarchical entropy according to claim 1, characterized in that: The implementation process of step (4) is as follows: Using decomposition operator Decompose the synchronous vibration data X to the nth level, and take the number of synchronous sampling points M in each tooth cycle of the milling cutter as the interval to obtain the low-frequency component and high-frequency component of the nth level, which is 2 n Each component is considered a node and marked as X n,e , The n-layer decomposition yields decomposition components.
6. The milling chatter feature extraction method based on synchronous hierarchical entropy according to claim 1 is characterized in that: The implementation process of step (5) is as follows: Calculate the synchronous hierarchical entropy of each node obtained by each layer decomposition; for the nth layer node X n,e Synchronous hierarchical entropy SHEn n,e The calculation process is as follows: first, the time delay is set to the time to obtain a data, the embedding dimension is set to 2, and the n-th layer node X is reconstructed. n,e The phase space of n,e ; Then calculate the phase space X′ n,e Two adjacent vectors v j and v j+1 Cosine similarity of as follows: And j =in j ·in j+1 , B j =||v j ||, v j =[x n,e (j),x n,e (j+1)], where j = 1, 2, 3, ..., (kL-n)M-1, v j is the jth vector in phase space, x n,e (j) is the nth layer node X n,e The jth element of ; Then divide the interval [-1,1] into ε intervals (I1,I2,…I ε ), calculate the above cosine similarity respectively The probability of being in each interval in, According to the probability of each interval Calculate the synchronous hierarchical entropy SHEn n,e , which is calculated as follows: Finally, for the n-layer decomposition decomposition components, we can get Synchronous level entropy.
7. The milling chatter feature extraction method based on synchronous hierarchical entropy according to claim 1 is characterized in that: The implementation process of step (6) is as follows: Calculate the result from step (5) The synchronous level entropy is regarded as the chatter feature, thus obtaining A vibration characteristic.