- Email: [email protected]
PVDF sensor based on-line mode coupling chatter detection in the boring process
Manufacturing Letters 16 (2018) 40–43
Contents lists available at ScienceDirect
Manufacturing Letters journal homepage: www.elsevier.com/locate/mfglet
PVDF sensor based on-line mode coupling chatter detection in the boring process Vinh Nguyen a,⇑, Shreyes Melkote a, Amar Deshamudre b, Maneesh Khanna b a b
George W. Woodruff School of Mechanical Engineering, Georgia Institute of Technology, Atlanta, GA, USA Siemens Energy, Inc., Charlotte, NC, USA
a r t i c l e
i n f o
Article history: Received 10 January 2018 Received in revised form 4 April 2018 Accepted 21 April 2018 Available online 23 April 2018 Keywords: Piezoelectric Cutting torque Vibrations Chatter detection Boring
a b s t r a c t Dynamic instability in the form of chatter vibrations during boring is detrimental to machining process. Therefore, an on-line chatter detection system is required to detect chatter before damage occurs. This paper evaluates a Polyvinylidene Fluoride (PVDF) piezoelectric thin-film based strain sensing system to detect mode coupling chatter in the boring process. In addition, this paper compares the performance of time domain based (autocorrelation), time–frequency based (wavelet transform), and frequency based (Fast Fourier Transform) methods for detecting mode coupling chatter. The PVDF sensor system is demonstrated to be a low-cost solution for detecting chatter in the boring process. Ó 2018 Published by Elsevier Ltd on behalf of Society of Manufacturing Engineers (SME).
1. Introduction During machining of gas turbine parts, dynamic instabilities such as chatter vibrations can negatively impact part quality, especially during boring. Therefore, an appropriate sensor-based process monitoring system is required to compensate for process modelling limitations. The majority of literature on chatter detection and suppression in machining is focused on the regenerative chatter mechanism [1–6], which occurs commonly in turning and milling operations performed on relatively rigid machine tools. However, the chatter mechanism in the boring process is primarily due to the difference in orientation between the two principal axes of stiffness and the resultant force vector [7,8]. This chatter mechanism, known as mode coupling chatter, occurs in symmetric low stiffness machining systems including long boring bars and serial link-based robotic milling [9]. While some work on mode coupling chatter modeling and ways to suppress it have been reported [10,11], methods for real-time detection of chatter in boring operations are lacking. This paper seeks to evaluate a low-cost Polyvinylidene Fluoride (PVDF) thin film wireless sensor system in detecting mode coupling chatter in the boring process. In addition, various chatter detection algorithms suitable for implementation in an embedded system are evaluated for monitoring of mode coupling chatter in ⇑ Corresponding author. E-mail address: [email protected] (V. Nguyen).
boring operations. The embedded chatter algorithms developed for detection of regenerative chatter in turning presented in [12] are extended to the boring process in this paper. Note that due to the differences between the underlying chatter mechanisms, the performance of the chatter detection system is expected to differ, thus contributing to the novel findings discussed in this paper. 2. Methodology This section describes the mechanism of chatter in boring. In addition, an overview of the PVDF thin film sensor-based chatter monitoring system hardware used to implement and evaluate the boring chatter algorithms is presented. 2.1. Mode coupling chatter mechanism For mode coupling stability analysis, a schematic of the boring process, as viewed from the feed direction, is shown in Fig. 1. Note that the cutting force is decomposed in the X and Y directions of a rotating frame via the angle a attached to the workpiece. The corresponding two degree of freedom dynamic system equation without considering damping is:
€x K px cos a K py cos a x x ½M þ ½K ¼ € K px sin a K py sin a y y y
ð1Þ
where [M] is the effective dynamic mass, [K] is the dynamic stiffness of the boring bar, and Kpx and Kpy are defined as the linear
https://doi.org/10.1016/j.mfglet.2018.04.004 2213-8463/Ó 2018 Published by Elsevier Ltd on behalf of Society of Manufacturing Engineers (SME).
41
V. Nguyen et al. / Manufacturing Letters 16 (2018) 40–43
½A ¼ ½VT
K px cos a K py cos a K max ½V K px sin a K py sin a 0
0 K min
ð3Þ
where the stability and oscillation frequencies depend on the eigenvalues of [A]. Note that the eigenvalues, and therefore their oscillatory terms, do not depend on spindle rotational velocity. The oscillatory frequency for mode coupling chatter has been shown to be near the system’s lowest natural frequency mode [10,11]. Hence, the hardware and algorithm for detecting mode coupling chatter must be capable of discerning the natural frequencies of mode coupling susceptible systems including boring bars, which are lower than frequencies encountered in regenerative chatter. Therefore, the robustness requirements of the chatter detection sensors and algorithms differ between chatter mechanisms. 2.2. PVDF based chatter monitoring system A schematic of the PVDF sensor based chatter monitoring system for boring is shown in Fig. 2. Specifically, the sensing and monitoring system components utilized in this work are as follows:
Fig. 1. Schematic of the boring process dynamic model.
sensitivities of cutting forces due to deflections in the X and Y directions (cutting coefficient), which depend on the cutting conditions and the workpiece material. Note that the machining force in the feed direction was not considered because the stiffness in the feed direction is much higher than the stiffness of the coupled modes in the X and Y directions. Also note that since the presence of damping generally increases the system stability, the undamped model in Eq. (1) yields a conservative solution for system stability. Because [M] and [K] are symmetric and semi-positive definite matrices they can be diagonalized by a transformation matrix [V], which results in the following after solving for the acceleration vector [9]:
€x x ¼ ½A € y y
ð2Þ
A PVDF-based sensor rosette capable of measuring the dynamic shear strains produced in the host structure (boring bar) during boring is used. The PVDF polymer can be laminated onto a sheet of polyester, resulting in a very thin sensor film (40 lm) thus minimizing its impact on the host structure’s dynamics [13]. The sensitivity and capability of the PVDF sensor rosette to measure the dynamic cutting torque signal in boring and its comparison to a quartz-based cutting force dynamometer was demonstrated in earlier work [14]. Two PVDF thin film sensors are mounted 45° from the vertical axis of the boring bar to isolate the dynamic torsional strains (from the bending and axial strains), from which the dynamic cutting torque signal
Fig. 2. Schematic (left) and tool path for boring experiment (right).
42
V. Nguyen et al. / Manufacturing Letters 16 (2018) 40–43
can be derived. The charge amplifier low frequency cutoff was configured to be 7.24 Hz while the sampling rate was set to 13 kHz. The benchmark microprocessor unit used to implement the boring chatter detection algorithms was a Arduino-based 72 MHz clock speed microcontroller (MK20DX256VLH7) with 256 kB memory and a 16-bit analog to digital conversion resolution [15]. Note that this microprocessor can be configured into embedded systems and is relatively inexpensive ($20). The chatter detection algorithms are embedded into the microprocessor to process the data in real-time. The computational requirements for the receiver unit can be lowered if the microcontroller unit can process the data and transmit when chatter is detected. Due to the microcontroller’s speed and memory constraints, the algorithm must be computationally efficient.
3. Boring experiments The chatter detection algorithms evaluated in this paper were based on the autocorrelation function, wavelet transform, and Fast Fourier Transform, respectively. These methods were determined to be suitable for microcontroller requirements due to their low memory and computational requirements. Specifics of the algorithms are as follows: The 1st autocorrelation coefficient (a1) was calculated every 1000 points. When mode coupling chatter develops, the dynamic torque data becomes less random. Thus, the 1st autocorrelation coefficient approaches 1 and therefore a threshold can be set.
Table 1 Boring Test Conditions. Test No.
RPM
Feed (mm/rev)
Radial depth of cut (mm)
Chatter frequency (Hz)
1 2 3 4 5 6 7 8 9 10 11 12 13
540 540 540 540 520 520 520 520 520 650 650 594 594
0.167 0.167 0.167 0.167 0.404 0.404 0.404 0.404 0.404 0.323 0.323 0.118 0.118
0.635 0.635 0.191 0.191 0.191 0.191 0.191 0.191 0.191 0.635 0.635 0.152 0.152
271 270.3 260 257.8 257.8 No Chatter No Chatter No Chatter No Chatter 265.5 254 No Chatter No Chatter
Fig. 3. Algorithm Comparison for Test 2.
V. Nguyen et al. / Manufacturing Letters 16 (2018) 40–43
The Second Generation Wavelet Transform (SGWT) is used to decompose the time domain signal into wavelet decompositions. The specific wavelet used in this work is the Haar-Wavelet [16]. To examine the appropriate decomposition, six (6) detail coefficients were computed and each level’s sensitivity to chatter was evaluated. A ratio of the amplitudes of the two largest FFT magnitudes was calculated to determine if a frequency is becoming dominant. As chatter arises, the ratio of the peaks (PR) in the FFT will increase as the power at one of the two frequencies becomes large.
43
was shown to be capable of detecting mode coupling chatter in boring operations. In addition, three proposed chatter detection algorithms suitable for low-cost on-line chatter monitoring were evaluated for mode coupling chatter detection in boring. In addition to being robust to false alarms, the most efficient chatter detection method was found to be the 1st autocorrelation coefficient. Thus, for boring processes where cutter runout is absent, the 1st autocorrelation function is recommended due to its low computational complexity. Acknowledgment
All tests were performed with a single-insert, 65 mm diameter boring bar (ISCAR BHFI MB16-MB50 with CAT-40 holder) with a CAT-40 to CAT-50 adapter. The tests were performed on a boring machine (Liné Machine Tools, Vegamill TF218). The test conditions are summarized in Table 1. Analysis of the experimental data reveals that the spindle speed, depth of cut, and feed do not significantly influence the chatter frequency, thus demonstrating that mode coupling chatter is the dominant dynamic instability mechanism. In these tests, the average chatter frequency was 262 Hz with a standard deviation of 6.64 Hz. Note that the demonstration of real-time chatter avoidance was not possible due to restricted access to the internal control architecture of the Siemens 840D controller used in this work. 4. Boring chatter results Fig. 3 shows the comparison of the three algorithms under examination. Note that in all cases, the PVDF sensor system was able to detect chatter. In contrast to the turning results in [12], the SGWT does not consistently produce a decomposition with the highest sensitivity. However, the 1st Autocorrelation coefficient and the FFT ratio are shown to approach chatter in a timely manner. Note that the 1st Autocorrelation coefficient did not trigger any false alarms when the cut was stable. This is the case for boring operations because the effect of cutter runout is not present following a previous hole drilling/boring pass. Thus, the 1st Autocorrelation Coefficient is more robust than in the turning process [12] while enjoying the computational efficiency advantages over the other algorithms evaluated in this paper. Note that, in addition to the computational and memory requirements of the FFT algorithm, a different algorithm threshold was needed for the boring process as opposed to turning [12] due to the different chatter mechanism in effect. 5. Conclusion A low-cost Polyvinylidene Fluoride (PVDF) thin film wireless sensor system developed for measuring the dynamic torque signal
This work was supported by Siemens Energy. References [1] Merdol S, Altintas Y. Multi frequency solution of chatter stability for low immersion milling. J Manuf Sci Eng 2004;126:459–66. [2] Rao BC, Shin YC. A comprehensive dynamic cutting force model for chatter prediction in turning. Int J Mach Tools Manuf 1999;39:1631–54. [3] Ma L, Melkote SN, Castle JB. A model-based computationally efficient method for on-line detection of chatter in milling. J Manuf Sci Eng 2013;135:031007. [4] Bao S-F, Zhang W-G, Yu J-Y, Qiao S-M, Yang F-L. A new approach to the early prediction of turning chatter. J Vib Acoust 1994;116:485–8. [5] Cardi AA, Firpi HA, Bement MT, Liang SY. Workpiece dynamic analysis and prediction during chatter of turning process. Mech Syst Sig Process 2008;22:1481–94. [6] Qian S, Sun Y, Xiong Z. Intelligent chatter detection based on wavelet packet node energy and LSSVM-RFE, 2015 IEEE International Conference on Advanced Intelligent Mechatronics (AIM). IEEE 2015:1514–9. [7] Pratt Jr JR. Vibration control for chatter suppression with application to boring bars. Virginia Polytechnic Institute and State University, Department of Engineering Science and Mechanics; 1997. [8] Stone B. Extension of chatter theory, chatter and machine tools. Springer; 2014. p. 27–55. [9] Pan Z, Zhang H, Zhu Z, Wang J. Chatter analysis of robotic machining process. J Mater Process Technol 2006;173:301–9. [10] Gasparetto A. A system theory approach to mode coupling chatter in machining. J Dyn Syst Meas Contr 1998;120:545–7. [11] Pan Z, Zhang H. Analysis and suppression of chatter in robotic machining process, International conference on control, automation and systems. IEEE 2007:595–600. [12] Nguyen V, Melkote S, Deshamudre A, Khanna M, Walker D. Comparison of online chatter detection algorithms in turning, International Symposium on Flexible Automation. Cleveland, OH: IEEE; 2016. p. 87–94. [13] Mohammadi B, Yousefi AA, Bellah SM. Effect of tensile strain rate and elongation on crystalline structure and piezoelectric properties of PVDF thin films. Polym Test 2007;26:42–50. [14] Nguyen V, Melkote S, Deshamudre A, Khanna M, Walker D. PVDF sensor based monitoring of single-point cutting. J Manuf Processes 2016. [15] I. Freescale Semiconductor, K20 Sub-Family Reference Manual, 2012. [16] Rein S, Reisslein M. Low-memory wavelet transforms for wireless sensor networks: a tutorial. IEEE Commun Surv Tutorials 2011;13:291–307.
Contents lists available at ScienceDirect
Manufacturing Letters journal homepage: www.elsevier.com/locate/mfglet
PVDF sensor based on-line mode coupling chatter detection in the boring process Vinh Nguyen a,⇑, Shreyes Melkote a, Amar Deshamudre b, Maneesh Khanna b a b
George W. Woodruff School of Mechanical Engineering, Georgia Institute of Technology, Atlanta, GA, USA Siemens Energy, Inc., Charlotte, NC, USA
a r t i c l e
i n f o
Article history: Received 10 January 2018 Received in revised form 4 April 2018 Accepted 21 April 2018 Available online 23 April 2018 Keywords: Piezoelectric Cutting torque Vibrations Chatter detection Boring
a b s t r a c t Dynamic instability in the form of chatter vibrations during boring is detrimental to machining process. Therefore, an on-line chatter detection system is required to detect chatter before damage occurs. This paper evaluates a Polyvinylidene Fluoride (PVDF) piezoelectric thin-film based strain sensing system to detect mode coupling chatter in the boring process. In addition, this paper compares the performance of time domain based (autocorrelation), time–frequency based (wavelet transform), and frequency based (Fast Fourier Transform) methods for detecting mode coupling chatter. The PVDF sensor system is demonstrated to be a low-cost solution for detecting chatter in the boring process. Ó 2018 Published by Elsevier Ltd on behalf of Society of Manufacturing Engineers (SME).
1. Introduction During machining of gas turbine parts, dynamic instabilities such as chatter vibrations can negatively impact part quality, especially during boring. Therefore, an appropriate sensor-based process monitoring system is required to compensate for process modelling limitations. The majority of literature on chatter detection and suppression in machining is focused on the regenerative chatter mechanism [1–6], which occurs commonly in turning and milling operations performed on relatively rigid machine tools. However, the chatter mechanism in the boring process is primarily due to the difference in orientation between the two principal axes of stiffness and the resultant force vector [7,8]. This chatter mechanism, known as mode coupling chatter, occurs in symmetric low stiffness machining systems including long boring bars and serial link-based robotic milling [9]. While some work on mode coupling chatter modeling and ways to suppress it have been reported [10,11], methods for real-time detection of chatter in boring operations are lacking. This paper seeks to evaluate a low-cost Polyvinylidene Fluoride (PVDF) thin film wireless sensor system in detecting mode coupling chatter in the boring process. In addition, various chatter detection algorithms suitable for implementation in an embedded system are evaluated for monitoring of mode coupling chatter in ⇑ Corresponding author. E-mail address: [email protected] (V. Nguyen).
boring operations. The embedded chatter algorithms developed for detection of regenerative chatter in turning presented in [12] are extended to the boring process in this paper. Note that due to the differences between the underlying chatter mechanisms, the performance of the chatter detection system is expected to differ, thus contributing to the novel findings discussed in this paper. 2. Methodology This section describes the mechanism of chatter in boring. In addition, an overview of the PVDF thin film sensor-based chatter monitoring system hardware used to implement and evaluate the boring chatter algorithms is presented. 2.1. Mode coupling chatter mechanism For mode coupling stability analysis, a schematic of the boring process, as viewed from the feed direction, is shown in Fig. 1. Note that the cutting force is decomposed in the X and Y directions of a rotating frame via the angle a attached to the workpiece. The corresponding two degree of freedom dynamic system equation without considering damping is:
€x K px cos a K py cos a x x ½M þ ½K ¼ € K px sin a K py sin a y y y
ð1Þ
where [M] is the effective dynamic mass, [K] is the dynamic stiffness of the boring bar, and Kpx and Kpy are defined as the linear
https://doi.org/10.1016/j.mfglet.2018.04.004 2213-8463/Ó 2018 Published by Elsevier Ltd on behalf of Society of Manufacturing Engineers (SME).
41
V. Nguyen et al. / Manufacturing Letters 16 (2018) 40–43
½A ¼ ½VT
K px cos a K py cos a K max ½V K px sin a K py sin a 0
0 K min
ð3Þ
where the stability and oscillation frequencies depend on the eigenvalues of [A]. Note that the eigenvalues, and therefore their oscillatory terms, do not depend on spindle rotational velocity. The oscillatory frequency for mode coupling chatter has been shown to be near the system’s lowest natural frequency mode [10,11]. Hence, the hardware and algorithm for detecting mode coupling chatter must be capable of discerning the natural frequencies of mode coupling susceptible systems including boring bars, which are lower than frequencies encountered in regenerative chatter. Therefore, the robustness requirements of the chatter detection sensors and algorithms differ between chatter mechanisms. 2.2. PVDF based chatter monitoring system A schematic of the PVDF sensor based chatter monitoring system for boring is shown in Fig. 2. Specifically, the sensing and monitoring system components utilized in this work are as follows:
Fig. 1. Schematic of the boring process dynamic model.
sensitivities of cutting forces due to deflections in the X and Y directions (cutting coefficient), which depend on the cutting conditions and the workpiece material. Note that the machining force in the feed direction was not considered because the stiffness in the feed direction is much higher than the stiffness of the coupled modes in the X and Y directions. Also note that since the presence of damping generally increases the system stability, the undamped model in Eq. (1) yields a conservative solution for system stability. Because [M] and [K] are symmetric and semi-positive definite matrices they can be diagonalized by a transformation matrix [V], which results in the following after solving for the acceleration vector [9]:
€x x ¼ ½A € y y
ð2Þ
A PVDF-based sensor rosette capable of measuring the dynamic shear strains produced in the host structure (boring bar) during boring is used. The PVDF polymer can be laminated onto a sheet of polyester, resulting in a very thin sensor film (40 lm) thus minimizing its impact on the host structure’s dynamics [13]. The sensitivity and capability of the PVDF sensor rosette to measure the dynamic cutting torque signal in boring and its comparison to a quartz-based cutting force dynamometer was demonstrated in earlier work [14]. Two PVDF thin film sensors are mounted 45° from the vertical axis of the boring bar to isolate the dynamic torsional strains (from the bending and axial strains), from which the dynamic cutting torque signal
Fig. 2. Schematic (left) and tool path for boring experiment (right).
42
V. Nguyen et al. / Manufacturing Letters 16 (2018) 40–43
can be derived. The charge amplifier low frequency cutoff was configured to be 7.24 Hz while the sampling rate was set to 13 kHz. The benchmark microprocessor unit used to implement the boring chatter detection algorithms was a Arduino-based 72 MHz clock speed microcontroller (MK20DX256VLH7) with 256 kB memory and a 16-bit analog to digital conversion resolution [15]. Note that this microprocessor can be configured into embedded systems and is relatively inexpensive ($20). The chatter detection algorithms are embedded into the microprocessor to process the data in real-time. The computational requirements for the receiver unit can be lowered if the microcontroller unit can process the data and transmit when chatter is detected. Due to the microcontroller’s speed and memory constraints, the algorithm must be computationally efficient.
3. Boring experiments The chatter detection algorithms evaluated in this paper were based on the autocorrelation function, wavelet transform, and Fast Fourier Transform, respectively. These methods were determined to be suitable for microcontroller requirements due to their low memory and computational requirements. Specifics of the algorithms are as follows: The 1st autocorrelation coefficient (a1) was calculated every 1000 points. When mode coupling chatter develops, the dynamic torque data becomes less random. Thus, the 1st autocorrelation coefficient approaches 1 and therefore a threshold can be set.
Table 1 Boring Test Conditions. Test No.
RPM
Feed (mm/rev)
Radial depth of cut (mm)
Chatter frequency (Hz)
1 2 3 4 5 6 7 8 9 10 11 12 13
540 540 540 540 520 520 520 520 520 650 650 594 594
0.167 0.167 0.167 0.167 0.404 0.404 0.404 0.404 0.404 0.323 0.323 0.118 0.118
0.635 0.635 0.191 0.191 0.191 0.191 0.191 0.191 0.191 0.635 0.635 0.152 0.152
271 270.3 260 257.8 257.8 No Chatter No Chatter No Chatter No Chatter 265.5 254 No Chatter No Chatter
Fig. 3. Algorithm Comparison for Test 2.
V. Nguyen et al. / Manufacturing Letters 16 (2018) 40–43
The Second Generation Wavelet Transform (SGWT) is used to decompose the time domain signal into wavelet decompositions. The specific wavelet used in this work is the Haar-Wavelet [16]. To examine the appropriate decomposition, six (6) detail coefficients were computed and each level’s sensitivity to chatter was evaluated. A ratio of the amplitudes of the two largest FFT magnitudes was calculated to determine if a frequency is becoming dominant. As chatter arises, the ratio of the peaks (PR) in the FFT will increase as the power at one of the two frequencies becomes large.
43
was shown to be capable of detecting mode coupling chatter in boring operations. In addition, three proposed chatter detection algorithms suitable for low-cost on-line chatter monitoring were evaluated for mode coupling chatter detection in boring. In addition to being robust to false alarms, the most efficient chatter detection method was found to be the 1st autocorrelation coefficient. Thus, for boring processes where cutter runout is absent, the 1st autocorrelation function is recommended due to its low computational complexity. Acknowledgment
All tests were performed with a single-insert, 65 mm diameter boring bar (ISCAR BHFI MB16-MB50 with CAT-40 holder) with a CAT-40 to CAT-50 adapter. The tests were performed on a boring machine (Liné Machine Tools, Vegamill TF218). The test conditions are summarized in Table 1. Analysis of the experimental data reveals that the spindle speed, depth of cut, and feed do not significantly influence the chatter frequency, thus demonstrating that mode coupling chatter is the dominant dynamic instability mechanism. In these tests, the average chatter frequency was 262 Hz with a standard deviation of 6.64 Hz. Note that the demonstration of real-time chatter avoidance was not possible due to restricted access to the internal control architecture of the Siemens 840D controller used in this work. 4. Boring chatter results Fig. 3 shows the comparison of the three algorithms under examination. Note that in all cases, the PVDF sensor system was able to detect chatter. In contrast to the turning results in [12], the SGWT does not consistently produce a decomposition with the highest sensitivity. However, the 1st Autocorrelation coefficient and the FFT ratio are shown to approach chatter in a timely manner. Note that the 1st Autocorrelation coefficient did not trigger any false alarms when the cut was stable. This is the case for boring operations because the effect of cutter runout is not present following a previous hole drilling/boring pass. Thus, the 1st Autocorrelation Coefficient is more robust than in the turning process [12] while enjoying the computational efficiency advantages over the other algorithms evaluated in this paper. Note that, in addition to the computational and memory requirements of the FFT algorithm, a different algorithm threshold was needed for the boring process as opposed to turning [12] due to the different chatter mechanism in effect. 5. Conclusion A low-cost Polyvinylidene Fluoride (PVDF) thin film wireless sensor system developed for measuring the dynamic torque signal
This work was supported by Siemens Energy. References [1] Merdol S, Altintas Y. Multi frequency solution of chatter stability for low immersion milling. J Manuf Sci Eng 2004;126:459–66. [2] Rao BC, Shin YC. A comprehensive dynamic cutting force model for chatter prediction in turning. Int J Mach Tools Manuf 1999;39:1631–54. [3] Ma L, Melkote SN, Castle JB. A model-based computationally efficient method for on-line detection of chatter in milling. J Manuf Sci Eng 2013;135:031007. [4] Bao S-F, Zhang W-G, Yu J-Y, Qiao S-M, Yang F-L. A new approach to the early prediction of turning chatter. J Vib Acoust 1994;116:485–8. [5] Cardi AA, Firpi HA, Bement MT, Liang SY. Workpiece dynamic analysis and prediction during chatter of turning process. Mech Syst Sig Process 2008;22:1481–94. [6] Qian S, Sun Y, Xiong Z. Intelligent chatter detection based on wavelet packet node energy and LSSVM-RFE, 2015 IEEE International Conference on Advanced Intelligent Mechatronics (AIM). IEEE 2015:1514–9. [7] Pratt Jr JR. Vibration control for chatter suppression with application to boring bars. Virginia Polytechnic Institute and State University, Department of Engineering Science and Mechanics; 1997. [8] Stone B. Extension of chatter theory, chatter and machine tools. Springer; 2014. p. 27–55. [9] Pan Z, Zhang H, Zhu Z, Wang J. Chatter analysis of robotic machining process. J Mater Process Technol 2006;173:301–9. [10] Gasparetto A. A system theory approach to mode coupling chatter in machining. J Dyn Syst Meas Contr 1998;120:545–7. [11] Pan Z, Zhang H. Analysis and suppression of chatter in robotic machining process, International conference on control, automation and systems. IEEE 2007:595–600. [12] Nguyen V, Melkote S, Deshamudre A, Khanna M, Walker D. Comparison of online chatter detection algorithms in turning, International Symposium on Flexible Automation. Cleveland, OH: IEEE; 2016. p. 87–94. [13] Mohammadi B, Yousefi AA, Bellah SM. Effect of tensile strain rate and elongation on crystalline structure and piezoelectric properties of PVDF thin films. Polym Test 2007;26:42–50. [14] Nguyen V, Melkote S, Deshamudre A, Khanna M, Walker D. PVDF sensor based monitoring of single-point cutting. J Manuf Processes 2016. [15] I. Freescale Semiconductor, K20 Sub-Family Reference Manual, 2012. [16] Rein S, Reisslein M. Low-memory wavelet transforms for wireless sensor networks: a tutorial. IEEE Commun Surv Tutorials 2011;13:291–307.