1. Introduction
Micro-milling is an important advanced manufacturing technology in the field of micromachining [
1]. It is the technology from electric processing to non-electric processing, silicon micro-processing to non-silicon micro-processing, and two-dimensional processing to three-dimensional processing. It has become an effective way to process small, complex 3D structural parts [
2,
3]. With the gradual expansion of processing object and application range, it is being used for more high-quality parts processing, such as in the aerospace, precision instruments, biomedical, automotive, and microelectronics fields. However, due to the huge variable cross-section, weak stiffness, and greatly reduced scale of the micro-milling tool, the processing mechanism of micro-milling is significantly different from traditional milling. It presents three typical characteristics [
4,
5]: ① The strains and positions of the micro-milling tools are different under different kinds and directions of complex impact forces. ② As the processing size is reduced to micron and sub-micron scales, the cutting depth of the micro-milling tool is often less than the grain diameter of the material, and many physical phenomena and basic laws in the macro world are no longer fully applicable to the micro world. The micro surface force, friction force, and heat transfer mode in the micro world play a major role. At the same time, the discontinuity and heterogeneity of materials also cause complex nonlinear cutting force fluctuation at high frequencies. ③ The ultra-high-speed spindle radial runout under the centrifugal force is amplified, caused by the inevitable installation errors of tool installation tilt and eccentricity. These extremely complex factors affect the variation of the cutting force and cutting vibration signal together, meanwhile forcing micro-milling tools to present a variety of motion patterns [
6]. In order to predict and suppress micro-milling tool vibration, two methods are commonly used: the vibration mechanical model and vibration measurement.
The vibration mechanical model is an important way to understand the generation and development of the vibration of micro-milling tools. Therefore, considering the scale effect and other factors, early researchers have performed many studies based on the conventional milling processing law [
2]. Huaizhong Li [
7] measured the vibration signals using accelerometers in the machining process and presented the characteristics related to chatter occurrence in micro-milling operations. Considering tool runout, the dynamic displacement of the tool, the workpiece contour left by previous tool paths, and the wavy surface, Wangqun Chen [
8] established an uncut chip thickness model. A deeper understanding of micro-milling processing rules, tool runout, the relative motion between the tool and the work piece, and the instantaneous uncut thickness change with the tool vibration were taken into account one after the other. David, C. [
9] evaluated the effect of runout on tool vibration based on experimental analysis under various cutting process parameters, such as the cutting speed, feed rate, and axial cutting depth. In order to improve the understanding of the connection between the dynamic cutting process and abnormal tool response, K.B. Mustapha [
10] presented a hybrid analytical model. The model combination of discrete and distributed structural elements was used for estimating the transverse response of the micro end mill. Based on the strain gradient elasticity theory and the Hamiltonian principle, Qinghua Song [
11] established a micro-milling tool model considering the shear deformation and rotational inertia of the micro-milling tool. Additionally, the static and dynamic characteristics of the micro-milling tool under different tool diameters and different slenderness ratios were studied. However, due to the high precision requirements of micro-milling, researchers have to take more complex factors that may cause micro-milling vibration into account to obtain a more accurate tool vibration prediction model. Xiaohong Lu [
12] established a micro-milling surface roughness model considering the tool vibration. The model studies the relationship between tool vibration and machined surface roughness. Then, another rotary Timoshenko beam model of the micro-milling spindle system was established, and the effects of spindle centrifugal force and gyro effect on the frequency response characteristics of the tool system and its stability region were studied [
13]. In summary, vibration, as an important factor affecting the stability of ultra-high speed precision micro-milling, is affected by a variety of complex factors, such as machine tool structure, material properties, processing parameters, tool runout, etc. Therefore, establishing a comprehensive model considering all factors to express the relationship between vibration and system parameters is very complex. Furthermore, the above research mostly focuses on the effects of tool vibration on work piece processing quality and lacks analysis of the tool vibration itself. Sensors and measurement techniques are an important complement to those research approaches.
Sensors and measurement techniques are other means for understanding tool vibration. Scholars use laser displacement sensors [
14,
15], acoustic emission sensors [
11], three-dimensional cutting force sensors [
16,
17], acceleration sensors [
2], or integrate the above-mentioned sensors [
18,
19,
20] to obtain signals of micro-milling vibration displacement, sound, cutting force, acceleration, current, and so on. They are used for monitoring, fault diagnosis, and the performance evaluation of machine tool or tool state, realizing function failure judgment, and processing the ability evaluation of machine tools. For example, Xiaohong Lu [
14] developed a micro-milling vibration measurement system based on a laser displacement sensor. The system can be extended to measure vibration in three directions simultaneously, which provides an experimental condition for the research of vibration suppression in micro-milling. Tsai, N.C. [
15], proposed a real-time flutter detection method based on acoustic emission and applied the acoustic feedback signal analysis and processing to control the spindle motor speed. In order to identify the chatter and tool wear simultaneously, Runqiong Wang [
17] proposed a multi-condition identification method based on sensor fusion by fusing sound, acceleration, and cutting bending moment signals data. Polli, M.L. [
20], comprehensively utilized force sensors, acoustic emission sensors, and acceleration sensors to obtain micro-milling processing signals. Nevertheless, these studies are more interested in the measurement itself. It is assumed that the measurement method obtained accurate tool vibration information but the vibration transfer law and vibration dissipation were not paid much attention.
Thus, the above research studied the vibration mechanical model, measurement method, and vibration prediction model of micro-milling tools from different angles. They undoubtedly deepen the general understanding of micro-milling and improve machining stability and accuracy. However, some questions currently remained unanswered; for instance, which vibration signals are reliable? What data are available in terms of flutter prediction? How should suitable measurement points to verify the model be selected. Furthermore, research on the vibration transfer characteristics of micro-milling tools is not enough, as their small size, large variable cross-section, and weak rigidity characteristics make them significantly different from traditional milling tools. Hence, this paper focuses on the above questions and tries to explore the vibration transfer law of micro-milling tools. Although there is much research on the vibration transfer characteristics of structures with variable cross-sections, they mainly focus on large structures [
21,
22,
23]. From the micro-milling tool vibration point of view, research is not sufficient, as micro-tools are of very small size (from the submicron to millimetre scale), with larger variable cross-sections (in a very short distances, the cross-sectional area of the tool tip and the tool bar differs by a hundred times) and weak rigidity characteristics. This is a different perspective and also the novelty of this paper compared with the existing research on micro-milling tool vibration.
In order to answer the questions mentioned above, this paper is organized into three sections. The first focuses on the vibration propagation model of the micro-milling tool. The two-dimensional vibration mechanics model and the vibration transfer matrix of the micro-milling tool are established. The second section is centred on the vibration characteristics analysis of the micro-milling tool. Two representative micro-milling tools are taken as an example. Their vibration propagation characteristics in the time domain and frequency domain are analysed, respectively. To verify the above model and simulation results, a micro-milling tool vibration measurement experimental system was set up, and a sensor array with four optical fibre displacement sensors was used to obtain the vibration displacements at different positions of the tool under pulse excitation. The third section presents the measurement results of two micro-milling tools in the machine start–stop process. The influence of the tool section area on vibration transmission is discussed. The three sections aim to explore the vibration transfer characteristics of micro-cutting tools in detail.
4. Experiment Results and Discussions
Figure 10 shows the micro-milling tool vibration measurement experimental system. It consists of seven main parts: a desktop milling machine (with micro-milling tool), a voltage-stabilized source, four light sources, an optical-electric conversion module, a DAQ card, and the optical fiber sensor array. The speed range of the desktop milling machines is from 0 to 10,000 RPM with the power of 480 W. It is made by the MENCHAO Company. The optical fiber sensor array consists of four optical fiber displacement sensors. The sensor [
30] was developed by the author with a measurement range of 2 mm and the sensitivity of 2.668 mv/µm. The maximum signal frequency of the sensor system designed by us is 50 kHz, its linear error is 0.12% with the max measurement uncertainty of 2.4 µm. The wavelength and power of the light source are 650 nm and 10 mW, respectively. An OPT01 chip is used in photoelectric conversion module, and its output voltage is approximately 0.45 V/µw at a 650 nm wavelength. The DAQ card USB-1901 made by ADLINK was chosen, which, with a maximum sampling rate of 250 kS/s USB 2.0-based high-performance DAQ modules, allows four different voltage input ranges.
The relative position of the sensor array and Tool A and Tool B is shown in
Figure 11. Sensor 1 is used for measuring the vibration displacement of the tool tip. Sensor 2, sensor 3, and sensor 4 are used to measure the vibration of the Tool A at 10 mm, 18 mm, and 24 mm, respectively, from the tool tip. Due to the limitations of Tool B and the size of the sensors, the vibration measurement points of Tool B of sensor 2, sensor 3, and sensor 4 are located at 8 mm, 20 mm, and 30 mm from the tool tip, respectively. In fact, in the machining process, the tool holder is the key interface connecting the tool and the machine tool. For example, the clamping rotary accuracy determines the dimensional accuracy of the workpiece and the vibration damping characteristics of the tool handle have a decisive impact on the service life of the tool and the installation error and runout of the tool on the tool handle also have a non-negligible impact on the tool vibration. However, considering the size of the micro-milling tool is very small compared with the tool holder, the vibration signal of the micro-milling tool is extremely weak, so the sensor layout is mainly extant around the tool.
In the experiments, one end of a very fine steel wire (diameter of 0.1 mm, 304 stainless steel) was hung on the tip of the tool, and the other end was fixed on a constant force spring sheet. The length of the constant force spring was adjusted. When the tension reached 4 N, the steel wire was cut quickly, and a transient force of 4 N was applied to the tip. Ten measurements per position were carried out. The impact vibration displacement response of Tool A is shown in
Figure 12.
Figure 12a shows the raw data of the impact response of Tool A in 5 ms, and the signal variation trend measured by sensor 1 agrees with the curve in
Figure 6. In the meantime, the vibration displacement signal measured by sensor 2, sensor 3, and sensor 4 decreases with the distance far away from the tip of Tool A, which is obvious. In order to show the difference between the signals of each sensor more clearly,
Figure 12b shows the first 0.5 ms impulse response signal (the signal is de-noised by the wavelet, using MATLAB toolbox). The proportion of vibration attenuation varies greatly compared with measurement point 1 (signal of sensor 1), measurement point 2 (signal of sensor 2), measurement point 3 (signal of sensor 3), and measurement point 4 (signal of sensor 4). The max vibration displacement of measurement point 1 is about 3.5 times measurement point 2, but nearly 18 times measurement point 3. Furthermore, the periodic vibration of measurement point 1 and measurement point 2 are highly similar; the vibration waveforms of measurement 3 and measurement 4 are obviously different to measurement point 1 and measurement point 2, based the time domain signals shown in
Figure 12. This result corresponds well with the simulation conclusion in
Figure 7a.
Figure 13 shows the frequency propagation of Tool A. Based on spectrum analysis of the received impact response signals of measurement point 1, point 2, point 3, and point 4 (corresponding to sensor 1 sensor 2, sensor 3, and sensor 4, respectively), it can be seen that there are two resonant frequencies of Tool A based on the signal received by sensor 1 (according to measurement point 1), 11.2 kHZ and 16.8 kHz. However, with the measurement points far away from the tool tip, the intensity of the signal with frequency of 11.2 kHz weakened. The 16.8 kHz signal suddenly disappeared at measurement point 2, measurement point 3, and measurement point 4. The spectrum analysis result of signal 4 (based on sensor 4) is more prominent, no distinct resonant frequencies appear in the signal.
Compared with the frequency propagation curves in
Figure 13 and the vibration displacement curves in
Figure 12, most information components are in measuring point 1. Although the same cross-section and similar signal waveforms are extant in the time domains of measurement point 1 and point 2, high frequency information is rapidly dissipated, moving from the tool tip to measuring point 2. It can be seen that this result is well consistent with the tool vibration information transmission characteristics in
Figure 4 and
Figure 5. Meanwhile, it can be seen that the high frequency vibration information of the tool tip is first prevented from propagating along the tool, as a large proportion of the tip area changes. However, the low-frequency vibration information of the tool dissipates relatively slowly in the propagation. Comparing with the results in
Figure 13 and
Figure 7b, it can be seen that the variation of the tool cross-section area not only affects the attenuation speed of the vibration signal amplitude but also has an obvious influence on the dissipation speed of the vibration signal frequency.
Figure 14 shows the impact response of Tool B at different positions.
Figure 14a shows the raw data of the impact response of Tool B in 5 ms, and
Figure 14b shows the first 0.5 ms impulse response signal (the signal is de-noised by the wavelet, using the MATLAB toolbox). The maximum vibration displacement of measurement point 1 is about 45 μm. However, the maximum vibration displacement of measurement point 2 is about 4 μm, and the overall intensity of the signal is relatively weak. The signal strength of measurement point 3 and point 4 are too weak to compare and analyze. Compared to
Figure 12, the influence of tool geometric shape is particularly significant. Compared with the distribution characteristics of the vibration attenuation factors of two micro-milling tools shown in
Figure 4, it can be seen that the variation of tool cross section area directly affects the vibration attenuation factor distribution, which leads to the transmission efficiency of tool vibration amplitude signal in the tool. This result agrees with the simulation conclusion in
Figure 9a.
Figure 15 shows the frequency propagation curves of Tool B based on impact response. Differently to Tool A, as can be seen in
Figure 13, there are three resonant frequencies of the tool based on the signal of measurement point 1 (sensor 1). They are 10.2 kHz, 17.6 kHz, and 26.7 Hz, respectively. On the other hand, the intensity of the signal based on sensor 2 (according to measurement point 2) is obviously weakened, which just leaves the frequency of 10.2 kHz and 17.6 kHz. The 26.7 kHz signal disappeared at measurement point 2. There are no distinct resonant frequencies that appear based on the spectrum analysis results of signal 3 (based on sensor 3 according to measurement point 3) and signal 4 (based on sensor 4 according to measurement 4) in
Figure 15. The information dissipates rapidly with the measurement points far away from the tip of Tool B. This means that the vibration displacement sensor must be placed as close to the tip as it can be, especially for micro-milling tools with large cross-sections. Compared to the results in
Figure 9b and
Figure 15, it can be seen that the amplitude and frequency of the vibration signal is mainly dissipated in the variable cross-section part of Tool B.
Comparing the time domain and frequency domain curves of Tool A and Tool B in
Figure 12,
Figure 13,
Figure 14 and
Figure 15, it can be seen that the vibration information propagation characteristics in the tool have obvious differences according to tool geometry.
Figure 16 shows the cross-section areas of Tool A and Tool B. As there are two parts of the cross-section of Tool A, the dissipative vibration process is mainly accomplished in two steps.
Figure 17 and
Figure 18 show the vibration propagation curves of Tool A and Tool B in the start–stop process. In Tool A, the start–stop process lasted for 5 s, then the tool speed was accelerated to 10,000 RPM in 1 s, was held for 2 s, and then slowed to 0 RPM in 2 s. In Tool B, the tool speed was accelerated to 10,000 RPM in 1 s, was held for 3 s, and then slowed to 0 RPM in 1 s. As
Figure 17 and
Figure 18 show, the attenuation of vibration displacement has a certain correlation with the change of tool section area. For Tool A, in the uniform rotation process, the vibration displacement of measuring point 1 is 7.3 µm, which is about 3.2 times the measurement of point 2. In fact, it is very difficult to measure the real tool tip vibration as its too small to find the correct measurement tool, so measuring point 1 actually has a certain distance of about 2.8 mm from the tool tip. The ratio of the cross-section of measurement point 2 and measurement point 1 is 4.5 and 3.2 divided by cross-section ratio 4.5 is 0.71. The vibration displacement of measuring point 2 is 6.2 times measurement point 3; however, the ratio of the cross-section of measurement point 3 and measurement point 2 is 16. Here, 6.2 divided by the cross-section ratio 16 is 0.39. For Tool B, the vibration displacement of measuring point 1 is 11.6 µm in the uniform rotation process, and it is about 38.7 times the measurement of point 2. The ratio of cross-section of measurement point 2 and measurement point 1 is 49.17, which can be calculated based on
Figure 16. Then, 38.7 divided by the cross-section ratio 49.17 is 0.79. Compared with Tool A, Tool B has a smaller geometry and a larger decay rate of the vibration amplitude. In contrast with the above results, it can be seen that a bigger ratio of the cross-section has a bigger attenuation of the vibration amplitude but a smaller geometry with a bigger decay rate of the vibration amplitude. Where the tool cross section area changes sharply, the tool vibration signal decreases faster with transmission and the high-frequency information dissipation speed is more obvious.
The above results show that the variation of tool cross section area has a significant effect on tool vibration transmission. Compared with the force dynamometer, the displacement sensor is more capable of chatter detection in high-speed micro-milling [
31]. However, without considering tool vibration transmission characteristics, a laser displacement sensor was mounted for tool bar vibration measurement and only the low-frequency information of tool vibration is obtained [
31]. To further demonstrate the results of this paper, we compared the experimental results with those of Reference [
32]. In order to analyze the device developed to perform the experimental modal analysis of micro-milling tools, impact tests of a cylindrical dummy tool and two micro-milling tools with different diameters (0.2 mm and 1 mm, respectively) were carried out to demonstrate the device application of Crichigno Filho, J.M. [
32]. The results of experimental FRF receptance and coherence are shown in
Figure 19. For the dummy tool, in
Figure 19a, although the signal intensity is attenuated, the information dissipation is not obvious. By contrast, in
Figure 19b, the consistency of vibration signal and excitation is obviously different at different positions of the tool. Which means that information dissipation occurs in the cutting tool during the transmission of the vibration signal. Although the structural parameters and excitation positions are different, this result still confirms the conclusion of this paper to some extent.
5. Conclusions
Focusing on the problem of micro-milling tool vibration propagation, a vibration propagation model was built for the micro end milling process. The vibration amplitude and frequency transfer characteristics of two typical micro-milling tools were analyzed. The simulation results show that the locations of the tool attenuation band are mainly distributed at the variable section of the tools. However, the stopband characteristics of two tools are significantly different from each other, as the section length and the section area ratio vary. A micro-milling tool vibration measurement experimental system was set up to test the above model. Pulse impact measurement experiments and start–stop vibration measurement experiments were carried out for Tool A and Tool B, respectively. The results show that a bigger ratio of the cross-section has a larger attenuation to vibration amplitude; meanwhile, smaller geometry with a bigger decay rate of vibration amplitude. However, limited by the small size of the micro-milling tool and the space for sensor placement, it is difficult to obtain the vibration data for the whole tool. In spite of this, the results of this paper still have reference significance: the information dissipates rapidly with the measurement points far away from the tip of the micro-milling tool. This means that the vibration displacement sensor must be placed as close to the tip as it can be, especially for micro-milling tools with large cross-sections.
The possible industrial applications for this research mainly include two aspects: (1) it provides a reference for tool geometry design, since the variation of cross-section area directly affects the transmission and the dissipation of tool vibration information, and (2) to guide the design and application of micro-milling tool vibration measurement sensors, because the complexity of the micro-milling tool structure and the precision of the cutting process makes it is very important to design a sensor and a detection system suitable for weak information acquisition [
33]. However, there are still some problems to be solved in the future, such as the effect of a tool holder, the machine parameters, and workpiece material on tool vibration. It is important to point out that micro-milling tools have typical characteristics of weak rigidity, and their vibrations include transverse vibration, longitudinal vibration, and torsional vibrations. The movement is a spatial concept. In this paper, only the transverse vibration propagation characteristics of the micro-milling tool were studied. In future research, the vibration propagation characteristics of the micro-milling tool will be analyzed from the perspective of three-dimensional space. Therefore, it will include two key issues: the first is to shed new light on the transmission and dissipation law of micro-milling tool vibration from an informatics perspective and establish a mathematical model that can unify physical tool structure, energy transfer, and information expression. The second is the dynamic precision measurement of the vibration hologram in the process of the micro-cutting tool, observing the spatial vibration pattern of micro-milling tool tip precisely.