Minqing Jing and Heng Liu

Transcription

1 Int. J. Mechatronics and Automation, Vol., No. 4, A reconfigurable system of chatter stability evaluation and monitoring for industrial machining processes Kaibo Lu* School of Mechanical Engineering, Xi an Jiaotong University, Xi an , Shaanxi Province, China and Division of Engineering and Physics, Wiles University, Wiles-Barre, PA 18766, USA aibo.lu@wiles.edu *Corresponding author Minqing Jing and Heng Liu School of Mechanical Engineering, Xi an Jiaotong University, Xi an , Shaanxi Province, China mjing@mail.xjtu.edu.cn hengliu@mail.xjtu.edu.cn Xiaoli Zhang Division of Engineering and Physics, Wiles University, Wiles-Barre, PA 18766, USA xiaoli.zhang@wiles.edu Abstract: This paper introduces an online prototype system including two independent modules of chatter stability prediction and monitoring for industrial applications. Turning processes are considered. At first, the stability charts are analysed by the Laplace transform. The charts obtained can assist the operators in selecting appropriate machining parameters before operations. Secondly, the characteristics of chatter are extracted in accordance with the variance and spectrum of vibration signals. The corresponding thresholds are adaptively confirmed by comparing the acquired signal with the stored stable cutting signal on line. Thirdly, the algorithms and modules introduced are implemented with the combination of the advantages of MATLAB and C sharp programming, in which delay strategy and overlap processing technique enable to improve the accuracy and performance of the system. Finally, industrial turning trials were conducted to verify the effectiveness of the proposed system. Keywords: chatter; stability; monitoring; hybrid programming; vibration; reconfiguration; signal processing. Reference to this paper should be made as follows: Lu, K., Jing, M., Liu, H. and Zhang, X. (01) A reconfigurable system of chatter stability evaluation and monitoring for industrial machining processes, Int. J. Mechatronics and Automation, Vol., No. 4, pp Biographical notes: Kaibo Lu is a PhD candidate in the Institute of Mechatronics and Information Systems at Xi an Jiaotong University. He received his BE in Vehicle Engineering from Hebei University of Engineering in 005 and MS in Mechanical Design and Theory at Taiyuan University of Technology in 008. His research interests include machine health monitoring and diagnosis, signal processing, dynamic analysis, and optimisation techniques. He is currently studying at Wiles University, PA, USA, as a joint PhD candidate from 011 to 01, which is supported by the China Scholarship Council. Minqing Jing is a Professor in the Institute of Mechatronics and Information Systems at Xi an Jiaotong University. He received his PhD in Mechanical Engineering from Xi an Jiaotong University. His research interests include Mechatronics, machine health monitoring and diagnosis, and electromagnetic levitation. Copyright 01 Inderscience Enterprises Ltd.

2 78 K. Lu et al. Heng Liu is an Associate Professor in the Institute of Mechatronics and Information Systems at Xi an Jiaotong University. He received his PhD in Mechanical Engineering at Xi an Jiaotong University in His research interests include dynamics of rotor-bearing systems, non-linear vibrations, and mechanical design. Xiaoli Zhang is an Assistant Professor in the Division of Engineering and Physics at Wiles University, PA, USA. She received her PhD in the Department of Biomedical Engineering, University of Nebrasa-Lincoln, Lincoln, NE, USA in 009, following her MS in 006 and BE in 003 both from the School of Mechanical Engineering at Xi an Jiaotong University, Xi an, China. Her research interests are robotics and medical device design and optimisation. 1 Introduction Machining processes play an important part in all the manufacturing processes. Usually, they are required to be further done for the parts manufactured by casting, forming, and various shaping processes., the undesired relative vibrations between the tool and the worpiece in machining (Tobias, 1965), results in poor surface finish, reduced productivity, and shortened tool life. In general, the following techniques or methods are widely used to ensure the quality of the production, that is, machining performance prediction and optimisation techniques (Muherjee and Ray, 006; Chandrasearan et al., 010) and process monitoring and control techniques (Altintas and Wec, 004; Liang et al., 004). In the aspect of machining process prediction and optimisation, heuristic or probabilistic computing tools, such as neural networs, fuzzy sets, genetic algorithms, simulated annealing, ant colony optimisation, particle swarm optimisation, as well as conventional mathematical programming methods, are used for predicting and optimising the performance of the machining processes. In the aspect of machining process monitoring and control, significant research efforts have been made on the development of online machining process monitoring and control systems, oriented towards multisensor data fusion, advanced signal processing techniques, and effective process control schemes. In this study, we adopt the theoretical analysis and online monitoring technique to suppress the chatter vibrations. stability and monitoring techniques have been broadly studied in the last decades. stability analysis is, in general, related to the achievement of the lobe curves that plot the boundary between stable and unstable regions with regard to spindle speed and width or depth of cut. The practical nowledge and suppression techniques of cutting chatter were introduced in Tetsutaro s (1984) boo. A thorough analysis of the dynamics of machining processes, and further research directions of major interest were presented by Chiriacescu (1990). stability in turning processes with varying spindle speed was investigated by three different methods (Insperger et al., 003), in which the semi-discretisation method has been proven most reliable and efficient as well as in predicting the stability of milling processes (Insperger and Stepan, 004; Gradise et al., 005). Shi and Tobias (1984) explained the basic non-linearity of finite amplitude instability experimentally and theoretically, concluding that the cutting-force non-linearity has the most important effect on the non-linear behaviour, and that to assume non-linear structural characteristic is unnecessary. Based on the results of Shi and Tobias, Stepan (001) studied the non-linear models of regenerative chatter in metal cutting, providing analytical explanation of the arising of subcritical Hopf bifurcation in Shi s experiments. The fundamental modelling and stability of chatter vibrations in machining processes were reviewed exclusively by Altintas and Wec (004). In the field of chatter monitoring, besides the conventional displacement, acceleration, and cutting force transducers, other inds of sensors were also applied to monitor the cutting conditions, such as acoustic emission sensors, power or current meters, and laser Doppler vibrometers. In addition to sensor hardware, signal processing techniques of wavelet analysis, expert systems, fuzzy logic, and artificial neutral networs have achieved maturity in recent years. Recently, Xie and Krishnan (011) introduced a method that combines discrete wavelet transform with principal component analysis for feature enhancement and signal decomposition in both spatial and temporal domains. An intelligent system was developed to monitor and optimise the ball-end milling processes by the use of force sensors (Cus et al., 006). Shi and Gindy (007) constructed an inprocess monitoring system mainly for broaching processes using strain sensors and wavelet analysis. Tangjitsitcharoen (009) investigated an online monitoring and detection of chip formation and cutting states for CNC turning by the use of force sensors and the Fourier transform algorithm. Even though cutting forces can reflect the cutting states most directly and sensitively, but the installation of force transducers is unfavourable to the original machine tool structure, and the force sensor is suffered from various factors in the industrial applications (Jia et al., 011). Tetsutaro (1984) and Chiriacescu (1990) also introduced the recognition of chatter vibrations distinguishing from the forced vibration in detail. The comprehensive reviews on the art of machining process monitoring can be found in Liang et al. (004). Although encouraging results on chatter stability and monitoring have been achieved, most of these results were achieved from the relatively ideal laboratory conditions. The deployment of these techniques in the worshop still presents a challenge as a balance between cost and effectiveness has to be struc by many machine shops (Stepan, 001; Liang et al., 004). In this study, based on an external reconfigurableintegrated monitoring and diagnosis prototype system of

3 A reconfigurable system of chatter stability evaluation and monitoring for industrial machining processes 79 computer numerical control (CNC), an online prototype system including two independent modules of chatter stability prediction and monitoring for industrial applications has been constructed cooperatively. Turning processes were considered. The stability charts of turning processes were evaluated analytically. The characteristics of chatter were extracted in the time and frequency domains. The corresponding algorithms and modules were implemented with hybrid programming. Finally, the extensive industrial machining trials were conducted to verify the effectiveness of the proposed techniques and system. This paper is organised in six sections. In Section, the chatter stability and characteristics of turning operations are investigated theoretically. Section 3 provides the specific software implementation technique. In Section 4, the industrial experiments and corresponding results are presented. The experimental results are discussed afterwards in Section 5. At last, Section 6 concludes this wor. cut a p are nominally constant. κ r and κ r, the geometry parameters of insert, represent the entering angle and the end cutting edge angle (ECEA), respectively. Figure 1 Regenerative mechanical model for cylindrical turning stability and characteristics The so-called regenerative effect has been the most accepted mechanism for machine tool chatter. This effect is referred to the instantaneous chip thicness undergoes both the inner modulation of a current cut and the outer modulation of the previous cut in amplitude. The corresponding mathematical model is expressed with delayed differential equations (DDEs). Turning belongs to the class of single-point machining operations where the process is continuous unlie the interrupted and periodic milling operations. The model of the turning is governed by an autonomous DDE. The mechanical model of cylindrical turning processes is depicted in Figure 1. The equation of motion has the form, bcωn x + ωζx n + ωnx = ( μx( x T) xt ( )) (1) where the natural frequency ω n of thes elastic structure, the damping ratio ζ, and the spring stiffness can be obtained via experimental or numerical modal analysis. b is the chip width. c is the equivalent cutting force coefficient in x direction normal to the cutting plane, which has considered the influence of the direction factor of the first principal mode. T is the time delay. The symbol μ is the overlap factor (Tetsutaro, 1984; Chiriacescu, 1990; Stepan, 001), which can be considered as the ratio between the amplitude of the cutting force due exclusively to the outer modulations and that caused only by the inner modulations of the chip thicness. In Figure 1, the overlap factor can be evaluated by, bo CD μ = b = i AB () where b o and b i represent the outer and the inner width, respectively. The advantage of this definition enables the estimation of μ for various machining operations and cutting conditions. In orthogonal turning or facing, μ = 1; in cylindrical turning, μ [0, 1]. The feedrate f and depth of The asymptotic stability of the equation of motion of turning processes in the Lyapunov sense is determined by the characteristic equation. By taing Laplace transform on both sides of equation (1) with x() t X( s) (3) t s (4) Equation (1) becomes bcωn s X() s + ωζsx () () st n s + ωnx s = ( μe 1) X() s (5) From equation (5), we obtain the characteristic equation of the system is bcωn s + ωζs st n + ωn ( μe 1) = 0 (6) Substituting pure imaginary characteristic roots s = jω into equation (6), where ω is the angular vibration frequency at the limit of the stability, and then, separating the real and imaginary parts yield the following equations of the stability lobes as b ωμ R( ) bcωn c n ω = ω + ωn + cos( ωt ) = 0 bcωμ n I( ω) = ωnζω+ sin( ωt ) = 0 (8) Solving equations (7) and (8), we obtain that the stability charts with respect to spindle speed and chip width can be expressed in a compact algebraic form, (7)

4 80 K. Lu et al. 60λωn N = i = 1,,..., K iπ+ θ ζλ blim = uc μsin θ where ζλ 1 λ ω = asin acos, μ ( ζλ) + ( ) ( ) ( 1 λ ζλ + 1 λ ) λ = ω/ ω. n The same results can be achieved by using the D-subdivision method (Kolmanovsii and Nosov, 1986). The stability charts obtained can assist the operators in selecting appropriate machining parameters before operations tae place. While if the used cutting parameters locate in unstable region divided by the critical stability chart, chatter vibrations may arise during machining processes. In this case, chatter monitoring function begins to tae effect. The cooperation of the stability prediction module and the monitoring module is illustrated in Figure. Figure Concept of the proposed system (see online version for colours) To identify chatter vibrations, the characteristic information relevant to chatter should be retrieved firstly. It has been proven that (Stepan, 001; Altintas and Wec, 004), when chatter occurs in turning processes, in the time domain, the amplitude of vibrations enlarges which can be observed by the increment of the signal s variance or other time-domain features quantitatively. In the frequency domain, the bandwidth of vibration signals transfers from broad to narrow as the energy distribution tends to concentrate when chatter arises. The critical chatter frequency is usually a little higher than the natural frequency of the most flexible machining structure. Accordingly, the norm of chatter arising in turning is formulated in accordance with the variance and spectrum of vibration signals. Note that this method can also be used in milling processes, which differs from turning operations only in the feature of the spectrum (Lu et al., 011b). Compared with other techniques mentioned in the introduction section, this strategy is faster and more effective, maing it more suitable for the industrial applications. (9) 3 Software implementation Based on the analytical results from the preceding section, a hybrid programming is implemented to develop the chatter stability prediction and monitoring modules in this section. It is well nown that MATLAB is a high-performance language for technical computing, and C# is a modern multi-paradigm programming language for developing software components deployed in distributed environments. So, the algorithms and modules are implemented with the combination of the advantages of MATLAB and C# programming. For the module of chatter stability prediction, at first, the algorithms were programmed into M-files by MATLAB language, and then MATLAB Compiler compiled these M- files into dll functions. These functions can be called by C# functions. The onset of chatter is a gradual process, and its threshold is different with machining conditions varying. In practice, the threshold value is empirically selected (Liang et al., 004). For the monitoring module, a relative threshold value is explored, which can improve the self-adaptiveness of the module. This module mainly contains four functions as follows, namely, the frequency extraction, new frequency search, variance calculation, and variance comparison functions. Initially, the compared library recording the normal variance and spectrum information is refreshed by the calculation of the stable cutting signals or previous records. Thereafter, the following newly acquired data are compared with this reference. Alarm is triggered off unless both the variance and spectrum satisfy the criteria quantitatively. To avoid the misjudgement due to the occasional impulse signals induced by the hard impure particle in worpiece or strong noises in the shop floor, a delay strategy, which the impulse signal occurs consecutive M (a configurable parameter) times will be considered as a chatter candidate, is employed. Besides, overlap processing technique is also applied, which can improve the accuracy of spectral estimation and help to update the display more quicly. Note that it is not true that the higher the overlap ratio the higher the spectral estimation. Figure 3 gives the scheme of the monitoring module. Figure 3 Bloc diagram of the chatter monitoring module After compiling the configuration files and the corresponding software interfaces, the modules designed for chatter prediction and monitoring, as one type of the modules, can be integrated into the reconfigurableintegrated monitoring and diagnostic prototype system mentioned above. Additionally, other potential modules for

5 A reconfigurable system of chatter stability evaluation and monitoring for industrial machining processes 81 monitoring and diagnosis of the functional components of CNC machines, such as spindle, power transmission, and actuation systems, etc., if necessary, can also be integrated in the future. That is very the meaning of the word of reconfigurable in this paper. Figure 4 Figure 5 Pictures taen in the experiments, (a) installation positions of the accelerometers (b) pinch turning (see online version for colours) (a) (b) Scheme of the industrial trials (see online version for colours) 4 Experiments and results Industrial trials were conducted to demonstrate the effectiveness of the developed system. Machining processes were performed on a nine-axis Turn-Mill Center of CHD5A with five-axis gang tooling system. The accelerometers of Dytran 355A6 were chosen to acquire the cutting conditions of the tools due to the ease of installation and lower cost, and mounted in the tangential and feed directions of cutter shan respectively, as shown in Figure 4(a). The data acquisition (DAQ), that is, the hardware of the external reconfigurable-integrated monitoring and diagnostic system, sampled the vibrations in continuous point sampling mode with sampling rate 1,04 Hz. A shaft from C45 steel was machined in the traditional single-point turning (ST) where the upper milling spindle was loced in a non-rotating position, and in the modern pinch turning (PT) where two opposing, ST tools impinged on the worpiece from opposite sides as shown in Figure 4(b), respectively. Facing was carried out as well. The scheme of the trial is shown in Figure 5. For the stability prediction, the following parameters were obtained by hammer test for calculation: ω n = 4 Hz, = N/m, and ζ = The equivalent cutting force coefficient was c = N/mm (Rao, 010). Both κ r and κ r are 7.5 degrees. The graphical representation of the projection of the restrictive conditions, such as the maximum power of the machine, tool life constraint, and cutting speed constraint, etc., onto the plane where the theoretical stability chart was plotted leads to a new image called the practical stability chart. In Figure 6, a practical stability chart mared in bold line for a rough turning is plotted, in which the main constraints are the minimum and maximum values of the spindle speed and that the cutting power should not exceed the nominal power P 0 of the electro-spindle motor. The stability curves can help the operators to choose or optimise the cutting conditions that are mostly dependent on chatterfree cutting, minimising machining costs, and maximising productivity before machining happens. The interface of the chatter monitoring module can be seen in Figure 7. The diagnostic results of the module were compared with the surface finish quality of the worpiece to judge if the diagnosis is reasonable. Besides, the diagnostic results could also be further verified by the numerical simulation. Figure 8 gives two typical examples of the unstable and stable cases, in which the adopted modal parameters are the same as those used in the experiment. It can be seen that the numerical results are in agreement with the stability analysis and the monitoring results. It should be noted that the chatter amplitude in the simulation is much larger than that in practical experiments. This is because the non-linear factor that the tool could leave the worpiece when the vibration enlarges to a certain extent (Shi and Tobias, 1984), is without considering in the numerical simulation. Testing 100 random vibration samples acquired from the trials, of which chatter data tae over 6%, the monitoring module can correctly confirm the cutting state at diagnostic success rate up to 87%. Table 1 gives some typical samples qualitatively, where the corresponding units of the cutting parameters are rpm, mm, and mm/min, respectively. After the careful observation and analysis, the misjudgement case was caused by the feedrate increased manually during the cutting. In fact, before the feedrate

6 8 K. Lu et al. increased the shape of that signal was smooth. After that, the amplitude increased slowly. To avoid this case, the reference database needs to be refreshed. In summary, the developed module has been proven effective. Figure 6 Interface of chatter stability prediction module (see online version for colours) Nmin Nmax Figure 7 Interface of chatter monitoring module (see online version for colours)

7 A reconfigurable system of chatter stability evaluation and monitoring for industrial machining processes 83 Figure 8 Numerical analysis (a) stable case (a p = 0.5 mm, N = 1,400 rpm) (b) unstable case (a p = 1.4 mm, N = 1,400 rpm) (see online version for colours) (a) (b) Table 1 Verification of chatter monitoring module Type Cutting parameters Signal Characteristics Recognise Turning 1400/0.5/14 (ST) Frequencies and variance stable. Stable 1400/1.4/14 (ST) New dominant frequencies near 37 Hz narrow; 7.35 times increase in variance. 1400\1.4\14 (ST) New dominant frequencies near 41 Hz narrow; 8.63 times increase in variance 1400/1.0/14 (ST) New dominant frequencies near 36 Hz narrow; 7.8 times increase in variance. 560/3.3/50 (Facing) New dominant frequencies near 334 Hz narrow; 5.8 times increase in variance. 800/.5/50 (Facing) New dominant frequencies near 350 Hz narrow; 4.8 times increase in variance. 110/.0/70 (Facing) 1400//14 56 (PT) New dominant frequencies near 17 Hz narrow; 6.7 times increase in variance. New dominant frequencies near 369 Hz narrow; 3.37 times increase in variance. Misjudgement 1400/.0/14 (PT) Frequencies and variance stable. Stable

8 84 K. Lu et al. 5 Discussions From Table 1, the dominant chatter frequencies are not a little greater than the natural frequency of the cutter as the theoretical analysis concludes. In addition, the chatter frequencies are different in different cutting passes even in the same type of machining. One reason for this could be the uncertain estimation of the parameters in the dynamic model. Additionally, in substance, the tool-worpiece cutting model is a time-variant non-linear dynamic system (Li et al., 01), especially in machining slender or thin worpieces (Chiriacescu, 1990), whereas a linear DDE was adopted for theoretical analysis in this paper. The chatter arising in the ST of the slender worpieces is related to the cutting pass and position. With the machining operations going, the length-to-diameter ratio of the worpiece is becoming greater, which leads to the low stiffness of the worpiece. And the different cutting position along the worpiece has the different stiffness (Lu et al., 011a), the varying stiffness may result in quasi-harmonic or self-excited vibrations in machining processes (Rivin, 1999). Through the experiments, it is found that the PT technique is much more advantageous than the ST to resist chatter or deformation of the worpiece. In the PT, the radial cutting force, leading to the dynamic chip thicness variation which substantially causes regenerative chatter vibrations, from the opposed cutter is cancelled out. Besides, the production rate is increased greatly. The issue of the lac of robust sensor hardware for machining condition monitoring is urgently required to be addressed. The harsh and complicated machining environment, combined with the inevitable presence of fluids, chips, and noises, maes rapid, accurate, and direct measurement difficult in many situations. Currently achievable levels of sensor robustness still cannot satisfy the requirements of industrial machining applications. 6 Conclusions stability and monitoring are of significance to suppress the unfavourable chatter vibrations in metal cutting. An online system consisting of the chatter stability analysis module and the chatter monitoring module has been constructed. The two modules enable to cooperate to control chatter vibrations effectively. The Laplace transform method was used to obtain the stability charts. Both variance and spectrum characteristics were considered as the criteria of chatter occurrence, and the corresponding threshold values were confirmed relatively in comparison with the reference database. Finally, the extensive industrial trials proved the techniques and system presented here effective. On the basis of the distinction in communication with the monitored targets, monitoring and diagnosis systems are usually classified into the external (loose-coupling) and the embedded (tight-coupling) layouts. The type employed in this paper belongs to the former, which has the advantages of high reliability, easy expansion and maintenance by virtue of the independent hardware from CNC; while which has the limitations of increased cost and troublesome sensor installation. The embedded layout somewhat reduces the stability of CNC, but its strengths in tight configuration and direct data communication with CNC provide another guide to the development of machining process monitoring systems. Future plans include optimising machining parameters with considering the chatter-free conditions in turning processes, and integrating the relevant optimisation algorithm as another module into the external reconfigurableintegrated monitoring and diagnostic prototype system for CNC machine tools. Acnowledgements This paper is a revised and expanded version of a paper entitled Industrial applications of chatter stability prediction and monitoring system for turning processes presented at 011 IEEE International Conference on Mechatronics and Automation (ICMA 011), which has been held at Beijing, China, from August 7 to August 10, 011. This wor has been supported in part by the National Key Technologies R&D Programme under Grant No. 01ZX , by the National Natural Science Foundation of China under Grants No and No , and by the Project No References Altintas, Y. and Wec, M. (004) stability of metal cutting and grinding, CIRP Annals Manufacturing Technology, Vol. 53, No., pp Chandrasearan, M., Muralidhar, M., Krishna, C. and Dixit, U.S. (010) Application of soft computing techniques in machining performance prediction and optimization: a literature review, The International Journal of Advanced Manufacturing Technology, Vol. 46, Nos. 6 8, pp Chiriacescu, S.T. (1990) Stability in the Dynamics of Metal Cutting, Elsevier Science Publishers, Oxford. Cus, F., Milfelner, M. and Balic, J. (006) An intelligent system for monitoring and optimization of ball-end milling process, Journal of Materials Processing Technology, Vol. 175, Nos. 1 3, pp Gradise, J., Kalveram, M., Insperger, T., Weinert, K., Stepan, G., Govear, E. and Grabec, I. (005) On stability prediction for milling, International Journal of Machine Tools and Manufacture, Vol. 45, Nos. 7 8, pp Insperger, T. and Stepan, G. (004) Updated semi-discretization method for periodic delay-differential equations with discrete delay, Journal of Numerical Methods in Engineering, Vol. 61, No. 1, pp

9 A reconfigurable system of chatter stability evaluation and monitoring for industrial machining processes 85 Insperger, T., Schmitz, T.L., Burns, T.J. and Stepan, G. (003) Comparison of analytical and numerical simulations for variable spindle speed turning, in Proceedings of IMECE 03, pp Jia, Z.Y., Liu, H.F., Wang, F.J. and Li, Q. (011) Output characteristics of giant magnetostrictive force sensor under the effect of various factors, International Journal of Mechatronics and Automation, Vol. 1, Nos. 3/4, pp Kolmanovsii, V.B. and Nosov, V.R. (1986) Stability of functional differential Equations, Academic Press, London. Li, C.H., Wang, S. and Ding, Y.C. (01) Effect of structural parameters of spindle system of ultra-high speed grinder on critical speed and vibration mode, International Journal of Mechatronics and Automation, Vol., No. 1, pp Liang, S.Y., Hecer, R.L. and Landers, R.G. (004) Machining process monitoring and control: the state-of-the-art, Journal of Manufacturing Science and Engineering, Vol. 16, No., pp Lu, K.B., Jing, M.Q., Liu, H. and Cong, C.X. (011a) Dynamic analysis of slender shaft in twin-spindle turning, Applied Mechanics and Materials, Vols , pp Lu, K.B., Jing, M.Q., Zhang, Y.Q. and Liu, H. (011b) Industrial applications of a system for chatter stability prediction and monitoring, in 011 ASME International Mechanical Engineering Congress and Exposition, Denver, Colorado, USA, (CD-ROM). Muherjee, I. and Ray, P.K. (006) A review of optimization techniques in metal cutting processes, Computer & Industrial Engineering, Vol. 50, No. 1, pp Rao, P.N. (010) Manufacturing Technology Volume Metal Cutting and Machine Tools, China Machine Press, Beijing. Rivin, E.I. (1999) Stiffness and Damping in Mechanical Design, CRC Press, New Yor. Shi, D. and Gindy, N.N. (007) Industry applications of online machining process monitoring system, IEEE/ASME Transactions on Mechatronics, Vol. 1, No. 5, pp Shi, H.M. and Tobias, S.A. (1984) Theory of finite amplitude machine tool instability, International Journal of Machine Tool Design Research, Vol. 4, No. 1, pp Stepan, G. (001) Modelling nonlinear regenerative effects in metal cutting, Transactions of the Royal Society of London A, Vol. 359, No. 1781, pp Tangjitsitcharoen, S. (009) In-process monitoring and detection of chip formation and chatter for CNC turning, Journal of Materials Processing Technology, Vol. 09, No. 10, pp Tetsutaro, H. (1984) Analysis and Control in Machining Processes, Shanghai Science and Technology Press, Shanghai. Tobias, S.A. (1965) Machine Tool Vibration, Blacie & Son Ltd., London. Xie, S.K. and Krishnan, S. (011) Signal classification via multi-scale PCA and empirical classification methods, International Journal of Mechatronics and Automation, Vol. 1, Nos. 3/4, pp.13 3.