A high-speed milling chatter detection method based on a small amount of measurement data

Abstract

This invention relates to the field of machining condition detection technology, and discloses a method for detecting chatter in high-speed milling based on a limited amount of measurement data. The method first determines the installation positions of the vibration sensor and speed sensor, and the amount of data collected per spindle revolution. n Data length for each time window nm and flutter threshold T Secondly, acquire periodically synchronized multi-point sampling data X; then, based on the amount of data collected per revolution... n The periodically synchronized multi-point sampling data is expanded into m columns of data Y; then data mining is performed on data Y, and the m columns of data are subtracted pairwise to obtain the mined m(m-1) columns of data Z; finally, the m(m-1) columns of data are calculated. Z The standard deviation of the data is used as a flutter indicator and compared with the flutter threshold T to determine whether flutter has occurred. This invention uses data mining technology to obtain more new data, shortens the time delay of flutter detection, and improves the reliability and real-time performance of flutter detection.

Description

Technical Field

[0001] This invention belongs to the field of condition detection and relates to a machining condition detection technology, specifically a chatter detection method based on a small amount of measurement data applied to high-speed milling or ultra-high-speed milling. Background Technology

[0002] High-speed or ultra-high-speed milling is a cutting-edge technology for the 21st century. Characterized by high efficiency, high precision, and high surface finish, it has found increasingly widespread application in industries such as automotive, aerospace, mold manufacturing, and instrumentation, yielding significant technical and economic benefits and becoming an important component of modern advanced manufacturing technology. During high-speed or ultra-high-speed milling, chatter is a common abnormal machining phenomenon. Chatter not only affects workpiece quality and processing efficiency but also accelerates tool wear and can even lead to machine tool damage. Therefore, chatter needs to be detected and controlled during milling operations.

[0003] Real-time milling chatter detection methods can effectively detect chatter occurrences. Currently, researchers and engineers have proposed various chatter detection techniques. However, existing chatter detection methods typically require the analysis of large amounts of data, which not only prolongs the chatter detection time but also requires significant storage space. Patent application number 201410620569.8 discloses a milling chatter detection method. This method first uses Comb filtering technology to remove the spindle speed frequency and its harmonic components from the chatter signal, eliminating the influence of normal vibration components on chatter detection; then, it uses C0 complexity and correlation coefficient as chatter detection indicators to identify whether chatter has occurred. Although this chatter detection method can effectively identify chatter, it cannot detect chatter generated during ultra-high-speed milling in real time and reliably. Patent application number 201910656987.5 discloses a chatter detection method based on signal processing and artificial intelligence. This chatter detection method first uses empirical mode decomposition to decompose the chatter signal into a set of intrinsic mode functions (IMA) signals. Then, it selects several IMA signals containing the most chatter information for analysis, extracting chatter features and constructing a multidimensional chatter feature observation space. Next, it uses a manifold learning algorithm to reduce the dimensionality of the multidimensional chatter feature observation space, and then fuses information from multiple chatter feature quantities. Finally, it uses a support vector machine to identify the chatter. Both of these chatter detection methods have relatively high computational costs and may not be suitable for chatter detection in high-speed or ultra-high-speed milling.

[0004] To reduce the computational cost and improve the speed of chatter detection, Schmitz et al. (Schmitz TL, Medicus K, Dutterer B. Exploring once-per-revolution audio signal variance as achatter indicator. Mach Sci Technol 2002; 6:215–33. http: / / dx.doi.org / 10.1081 / MST-120005957.) proposed using periodic synchronous single-point sampling technology to acquire chatter signals and utilize the variance of these signals to identify chatter. As the name suggests, periodic synchronous single-point sampling technology refers to achieving synchronous sampling once per revolution of the spindle, thereby obtaining a set of periodic synchronous single-point sampled data. Chatter occurring during milling can be considered as a bifurcation or chaotic phenomenon. Without considering the influence of noise, the periodic synchronous single-point sampled data for the chatter signal is a fixed value (or constant), while for the chatter signal itself, the periodic synchronous single-point sampled data is no longer a fixed value but exhibits fluctuations. Based on this characteristic of periodically synchronized single-point sampling data, we can identify chatter. Although this method can significantly reduce the amount of data collected and has lower computational costs, single-point synchronized sampling data cannot reliably display chatter information. Furthermore, since periodically synchronized single-point sampling technology only collects one data point per spindle revolution, to improve the reliability of chatter detection, data needs to be obtained over multiple spindle rotation cycles, increasing the detection time window and thus reducing the real-time performance of chatter detection.

[0005] To overcome the shortcomings of periodic synchronous single-point sampling technology in chatter detection, the invention patent with patent number ZL202210619937.1 collects multiple data points per spindle revolution, increasing the amount of data collected and fusing chatter information from multiple vibration sensors, thereby shortening data acquisition time and improving the real-time performance of chatter detection. While using multiple sensors to collect chatter data can yield more information, it increases the cost of chatter detection. Furthermore, using multiple sensors also increases the false alarm rate due to sensor failure.

[0006] The problem this invention aims to solve is how to quickly detect chatter when using a single vibration sensor to acquire limited data. To address this issue, this invention develops data mining technology based on the characteristics of chatter and discloses a high-speed milling chatter detection method based on a limited amount of measurement data. Summary of the Invention

[0007] Purpose of the invention: In view of the problems pointed out in the background art, the present invention discloses a high-speed milling chatter detection method based on a small amount of measurement data, which can quickly and reliably detect chatter when a small amount of measurement data is obtained by using a single vibration sensor.

[0008] Technical solution: This invention discloses a method for detecting chatter in high-speed milling based on a small amount of measurement data, specifically including the following steps:

[0009] Step 1: Determine the installation locations of the vibration sensor and speed sensor, the amount of data n collected per spindle revolution, the data length nm of each time window, and the chatter threshold T;

[0010] Step 2: Based on the information collected by the vibration sensor and the speed sensor, obtain periodic synchronous multi-point sampling data X;

[0011] Step 3: Based on the amount of data n collected per revolution, expand the periodically synchronized multi-point sampling data into m columns of data Y;

[0012] Step 4: Perform data mining on the expanded periodic synchronous multi-point sampling data. Subtract each pair of m columns of data to obtain the mined m(m-1) columns of data Z.

[0013] Step 5: Calculate the standard deviation σ of the m(m-1) columns of data Z after mining, and compare the standard deviation σ as the flutter index with the flutter threshold T; if the flutter index σ is greater than or equal to the predetermined flutter threshold T, it is determined that flutter has occurred; otherwise, it is considered that flutter has not occurred.

[0014] Furthermore, in step 1, the workpieces to be processed are divided into thin-walled workpieces and non-thin-walled workpieces; for milling of non-thin-walled workpieces, vibration sensors and speed sensors are installed on the spindle of a high-speed milling machine to obtain data; for milling of thin-walled workpieces, vibration sensors are installed on the thin-walled workpieces to obtain chatter vibration data, and speed sensors are installed on the spindle of a high-speed milling machine to obtain speed data.

[0015] Furthermore, the vibration sensor is an acceleration sensor, a velocity sensor, or a displacement sensor.

[0016] Furthermore, the acquired periodic synchronous multi-point sampling data is represented as X = [x(1), x(2), x(3), ..., x(n), ..., x(nm)], and the expanded data Y in step 3 is:

[0017]

[0018] Furthermore, in step 4, data mining is performed on the expanded periodic synchronous multi-point sampling data Y, and the data in the m columns are subtracted pairwise. The data after data mining is as follows:

[0019]

[0020] Beneficial effects:

[0021] 1. The method of this invention first considers the data within each rotation of the spindle as a set of data based on the characteristics of chatter. Then, it subtracts the displacement sampling data at the same rotation angle of the spindle between each set to extract more data. Finally, it determines whether chatter has occurred based on the standard deviation of the extracted data. This invention extracts more data reflecting chatter characteristics from a small amount of measurement data and uses the extracted data to detect chatter, thus improving the reliability of chatter detection.

[0022] 2. This invention expands the periodic synchronous multi-point sampling data to m columns based on the amount of data (n) collected per revolution. The purpose of expanding this periodic synchronous multi-point sampling data is to subsequently mine more data that can characterize flutter features. Compared to the periodic synchronous multi-point sampling data Y, the mined periodic synchronous multi-point sampling data Z contains more data and more flutter information. Therefore, the method of this invention can reduce the false diagnosis rate and false negative rate of flutter detection.

[0023] 3. The method of this invention detects chatter based on a small amount of measurement data, reducing data sampling costs and chatter detection delays. Furthermore, the method has low computational costs and short detection time delays, enabling rapid chatter diagnosis. Therefore, it can be applied to high-speed milling and ultra-high-speed milling systems.

[0024] 4. The method of the present invention has low computational complexity, and can use a control unit with slower computation speed, thereby reducing the cost of the control unit.

[0025] 5. The method of this invention employs periodic synchronous multi-point sampling technology, ensuring that the amount of synchronous data acquired per spindle revolution is unaffected by the spindle speed. Therefore, the method of this invention can be applied to variable speed ultra-high-speed milling systems. Attached Figure Description

[0026] Figure 1 This is a flowchart of the flutter detection method of the present invention;

[0027] Figure 2 This includes displacement data and periodic synchronous multi-point sampling data of thin-walled workpieces in the direction perpendicular to the milling process without chatter.

[0028] Figure 3 This includes displacement data and periodic synchronous multi-point sampling data of thin-walled workpieces in the direction perpendicular to milling when chattering occurs;

[0029] Figure 4 To implement the periodic synchronous multi-point sampling data of thin-walled workpieces in the direction perpendicular to the milling process without chattering after data mining;

[0030] Figure 5 To implement data mining, the periodic synchronous multi-point sampling data of thin-walled workpieces with chatter in the direction perpendicular to the milling process is obtained. Detailed Implementation

[0031] The implementation process of the method of the present invention will be described below with reference to the accompanying drawings.

[0032] In this embodiment, the workpiece being processed is a thin-walled workpiece, therefore the vibration sensor is mounted on the thin-walled workpiece.

[0033] refer to Figure 1 The present invention discloses a method for detecting chatter in high-speed milling based on a small amount of measurement data, comprising the following steps:

[0034] Step 1: Determine the amount of data collected per spindle revolution n=100, and the data length of each time window nm=500. It can be seen that a total of data was collected for each time window, which is the spindle rotation m=5 revolutions; specify the chatter threshold T as 10.

[0035] Workpieces can be categorized into thin-walled and non-thin-walled workpieces; this embodiment uses thin-walled workpieces. For milling thin-walled workpieces, a vibration sensor (which can be an acceleration sensor, velocity sensor, or displacement sensor) needs to be mounted on the workpiece to measure its displacement perpendicular to the machining direction. In this embodiment, a speed sensor will be mounted on the spindle of a high-speed milling machine to obtain speed data.

[0036] Step 2: Acquire periodic synchronous multi-point sampling data.

[0037] A vibration sensor mounted on the thin-walled workpiece measures the vibration displacement signal perpendicular to the machining direction. A speed sensor mounted on the spindle acquires spindle speed information. The vibration displacement signal without chatter and the periodically synchronized multi-point sampling data are shown below. Figure 2 As shown, the vibration displacement signal and the periodically synchronized multi-point sampling signal during flutter are respectively as follows: Figure 3 As shown.

[0038] Step 3: Expand the acquired periodic synchronous multi-point sampling data.

[0039] Since the time window length is set to nm=500 and n=100 data points are collected per revolution, the periodic synchronous multi-point sampling data can be expanded into 100 rows and 5 columns of data, with each column containing 100 data points.

[0040]

[0041] Step 4: Perform data mining on the expanded periodic synchronous multi-point sampling data.

[0042] Data mining was performed on the extended periodic synchronous multi-point sampling data Y. After subtracting each pair of the 5 data columns, the 20 data columns Z after data mining were obtained.

[0043]

[0044] After data mining, the periodic synchronous multi-point sampling data of the thin-walled workpiece in the direction perpendicular to the milling process without chatter is as follows: Figure 4 As shown, the data during flutter is as follows: Figure 5 As shown.

[0045] Step 5: Determine whether flutter has occurred.

[0046] Calculate the standard deviation σ of the data Z after data mining. The standard deviation σ without flutter is 6.1, which is less than the predetermined flutter threshold T = 10. The flutter detection program returns to step two to continue flutter detection until flutter is detected, or until the flutter detection process is terminated. For the periodically synchronized multi-point sampled data Z with flutter, the calculated standard deviation σ is 88.6, which is much greater than the predetermined flutter threshold T = 10, indicating that flutter has occurred.

[0047] The method of this invention has low computational complexity, thus enabling rapid chatter detection and applicability to high-speed and ultra-high-speed milling. Furthermore, this method groups the synchronous data for each revolution, subtracts the periodic synchronous sampling data from each group, and determines the presence of chatter based on the standard deviation of the data Z after data mining. When there is no chatter, the standard deviation of data Z is small, approaching 0 when there is no noise; while when chatter is present, the standard deviation of data Z is large.

[0048] Compared to traditional flutter detection, the method of this invention uses only a single vibration sensor and acquires a small amount of measurement data. Data mining techniques are then used to derive more data reflecting flutter information, significantly reducing the economic cost of using sensors. Simultaneously, it lowers the false alarm rate of flutter detection due to sensor failure. Furthermore, the method of this invention exhibits high sensitivity and reliability in flutter detection. The flutter index of traditional methods is 0.29 in the absence of flutter and 0.81 in the presence of flutter, showing a small difference. In contrast, the flutter index of the method of this invention is 6.1 in the absence of flutter and 88.6 in the presence of flutter, showing a larger difference and enabling more sensitive and reliable identification of early flutter. Therefore, the method of this invention can detect flutter sensitively and reliably.

[0049] The above embodiments are only for illustrating the technical concept and features of the present invention, and are intended to enable those skilled in the art to understand the content of the present invention and implement it accordingly. They should not be construed as limiting the scope of protection of the present invention. All equivalent transformations or modifications made in accordance with the spirit and essence of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A method for detecting chatter in high-speed milling based on a small amount of measurement data, characterized in that, Includes the following steps: Step 1: Determine the installation locations of the vibration sensor and speed sensor, the amount of data n collected per spindle revolution, the data length nm of each time window, and the chatter threshold T; Step 2: Based on the information collected by the vibration sensor and the speed sensor, obtain periodic synchronous multi-point sampling data X; Step 3: Based on the amount of data n collected per revolution, expand the periodically synchronized multi-point sampling data into m columns of data Y; Step 4: Perform data mining on the expanded periodic synchronous multi-point sampling data Y. Subtract each pair of m columns of data to obtain the mined m(m-1) columns of data Z. Step 5: Calculate the standard deviation σ of the m(m-1) columns of data Z after mining, and compare the standard deviation σ with the flutter index and the flutter threshold T. If the flutter index σ is greater than or equal to the predetermined flutter threshold T, it is determined that flutter has occurred; otherwise, it is considered that flutter has not occurred.

2. The high-speed milling chatter detection method based on a small amount of measurement data according to claim 1, characterized in that, In step 1, the workpieces are divided into thin-walled workpieces and non-thin-walled workpieces. For milling non-thin-walled workpieces, vibration sensors and speed sensors are installed on the spindle of a high-speed milling machine to obtain data. For milling thin-walled workpieces, vibration sensors are installed on the thin-walled workpieces to obtain vibration data, and speed sensors are installed on the spindle of a high-speed milling machine to obtain speed data.

3. The high-speed milling chatter detection method based on a small amount of measurement data according to claim 2, characterized in that, The vibration sensor is an acceleration sensor, a velocity sensor, or a displacement sensor.

4. The high-speed milling chatter detection method based on a small amount of measurement data according to claim 1, characterized in that, The acquired periodic synchronous multi-point sampling data is represented as X = [x(1), x(2), x(3), ..., x(n), ..., x(nm)], and the expanded data Y in step 3 is:

5. The high-speed milling chatter detection method based on a small amount of measurement data according to claim 1, characterized in that, In step 4, data mining is performed on the expanded periodic synchronous multi-point sampling data Y. Each of the m columns of data is subtracted from the other pairwise. The data after data mining is as follows:

Images