1: \documentclass{jsv}
2: \usepackage{amssyme,equation,bm}
3: \begin{document}
4: \draft
5: \volume{157}
6: \runtitle{VIBRATIONS IN A CUTTING PROCESS}
7: \runauthor{J. LIPSKI ET AL.}
8: \begin{frontmatter}
9: \title{SURFACE QUALITY OF A
10: WORK MATERIAL INFLUENCE ON
11: VIBRATIONS IN A CUTTING PROCESS}
12: \author{J. L{\sc IPSKI}, G. L{\sc ITAK},
13: R. R{\sc USINEK}, K.
14: S{\sc ZABELSKI}}
15: \author{A. T{\sc ETER}, J.
16: W{\sc ARMI\'NSKI} {\sc AND} K. Z{\sc ALESKI}}
17: \address{
18: Department of Applied Mechanics, Technical University of Lublin
19: Nadbystrzycka 36, PL-20-618 Lublin, Poland
20: }
21: %\received{$23\ October \ 2000$}
22: %\and \in \ final\ form {$\ 16\ March \ 2001$}
23: \received{$23\ October\ 2000,\ and\ in\ final\ form\ 21\ March\ 2001$}
24: \begin{abstract}
25: The problem of stability in the machining processes is an important task.
26: It is strictly connected with
27: the final quality of a product. In this paper we consider vibrations of a
28: tool-workpiece system in a
29: straight turning process induced by random disturbances and
30: their effect on a product surface. Basing on experimentally obtained
31: system parameters we have done the
32: simulations using one degree of freedom model. The noise has been
33: introduced to
34: the model by the Langevin equation.
35: We have also analyzed the product surface shape and its
36: dependence on the level of noise.
37: \end{abstract}
38: \end{frontmatter}
39:
40: \section{Introduction}
41: \noindent
42: The quality of a final surface in a cutting process is of a natural
43: interest of industry and technology.
44: Grabec \cite{1,2} and Gradisek with co-workers \cite{3} analyzed a simple
45: orthogonal
46: cutting model and found that,
47: chaotic conditions of tool-workpiece system,
48: given by appropriate system parameters, are possible. As they have demonstrated,
49: the appearance of such chaotic
50: conditions can have crucial effect on the stability of cutting process.
51: The chaotic vibrations also were
52: investigated experimentally by Tansel and others in \cite{4}.
53: On the other hand instabilities of cutting process have been known
54: for long time as a chatter phenomenon \cite{5,6,7}. The mechanism their appearance
55: includes the
56: nonlinear self-excitation in the cutting process, which
57: leads to vibrations
58: with larger value of amplitude, beyond the admissible limit. One of the source of
59: instabilities can be identified in the roughness of a material initial surface, which
60: introduces
61: randomness of
62: the material resistance during
63: the dynamic process. Wiercigroch and Cheng \cite{8} have investigated the
64: influence of noise on the orthogonal cutting system. In
65: their analysis they started from the spectral
66: representation of stochastic process. Nevertheless, the most common
67: treatment of dynamical processes
68: influenced by noise is the Fokker-Plank approach \cite{9,10,11}.
69: Because of some considerable difficulties which it meets by solving in higher dimensions
70: as well as for numerical reasons it could be transformed to the corresponding Langevin equations \cite{10,11}.
71: Here,
72: following the
73: papers \cite{11,12,13,14}, we use the stochastic Langevin equations with
74: an additive white noise and solve the dynamic
75: equations of examined system.
76: Previous articles \cite{8,11,13} devoted to cutting process
77: in
78: presence of noise, focused rather on the problem of
79: dynamics and possibilities of bifurcations induced by the random
80: disturbances.
81: Interestingly, Wiercigroch and Cheng [8] and later Przystupa and Litak [11], investigating
82: orthogonal cutting process with two degrees of freedom, claimed that in some conditions weak noise can even
83: stabilize the chaotic attractor.
84: On the other hand paper \cite{14} deals with the reconstruction of the system dynamics
85: from the stochastic time series. In their treatment the cutting process was assumed
86: to be deterministic but the measured data were influenced by noise
87: coming from the measurement procedure.
88:
89: Our paper is also a contribution
90: to a complicated problem of dynamics
91: of a cutting process, but we focus on the final quality of
92: a product surface, and in this context, in the
93: stability of process. Note, in Fig. 1, the shape of an initial
94: surface of a cut workpiece. The shape has numerous imperfections which
95: we will model by random deviations from the ideal cylindrical surface.
96: Adopting a
97: simple one degree-of-freedom model of regenerative cutting \cite{15}
98: we included the effect of a previous
99: pass of a straight turning process
100: by a time delay term \cite{16}.
101:
102: \section{Deterministic Model of Cutting process}
103:
104:
105: The physical model of a straight turning process, corresponding to our experimental system, is presented
106: in Fig. 2. Here we have introduced
107: the following notations: $v_f$ is a relative velocity between the tool and the workpiece, $h_0$ is an
108: assumed
109: initial while $h$ an actual cutting
110: depth; $w$ is a principal axis of relative vibrations;
111: $y$ indicates the direction normal to the axis of a workpiece
112: symmetry, $\kappa$ is a tool cutting edge angle;
113: $k$ and $c$ are the stiffness and damping of the system respectively; $n$ denotes
114: a rotational velocity of a workpiece;
115: $f$ denotes the direction of feed in a straight
116: turning; m is the effective mass of the system.
117:
118:
119: The main vibration in $w$ direction, perpendicular to the cutting edge, (Fig. 2) and to be precise we should
120: analyze vibrations as well as cutting force in $w$ direction.
121: However we are interested in the final surface profile given by the time history of $y$
122: and not by the actual profile $w$.
123: To analyze vibration in $y$
124: we have done simultaneous
125: projecting vibrations and forces into $y$ direction.
126: Thus the deterministic equation of motion of dynamical system, projected on a
127: normal (to final surface) direction $y$, can be written as
128: follows \cite{15}:
129: \begin{equation}
130: \ddot y + 2\tilde n \dot y + p^2 y = \frac{K}{m} g_y(h,v_f),
131: \end{equation}
132: where $p$ is a natural frequency of free vibrations of the workpiece $p^2=k/m$ while $\tilde n=
133: c/m$.
134: Nonlinearities, appearing in that system,
135: are included in the $g_y$ function
136: \cite{1,2,3,8,15}:
137: \begin{equation}
138: g_y(h,v_f)= \left[ c_2 \left( \left| \frac{v_f}{v_0} \right| - 1 \right)^2
139: +1 \right] \left[ c_3 \left( \frac{h}{h_0} -1 \right)^2 +1 \right]
140: \frac{h}{h_0} \Theta (h) {\rm Sgn} (v_f).
141: \end{equation}
142: Cutting depth $h$ and relative velocity $v_f$ are defined \cite{14,15}:
143: \begin{equation}
144: h=h_0+ y(t') -y(t),~~~~v_f=1-\frac{\dot y}{v_0}.
145: \end{equation}
146: $\Theta(h)$ and ${\rm Sgn} (v_f)$ correspond to step functions: Heaviside
147: and sign
148: functions respectively and $v_0$ ia the linear velocity of a rotational
149: motion of a workpiece during a steady cutting process.
150:
151: $t'$ is the time of
152: a previous pass:
153: \begin{equation}
154: t'=t-\Delta t,
155: \end{equation}
156: where $\Delta t$ is a workpiece revolution time during machining.
157: The shape of a nonlinear function $g_y$ (Eq. 2) dependent
158: on $h$ and $v_f$ is presented in Fig. 3. Note the two-dimensional
159: surface $g_y=g_y(h,v_f)$
160: was plotted only for positively defined $h$.
161: In case of negative $h$, the force on the left hand side of Eq. 1, is zero
162: because of the contact loss between
163: the tool and a workpiece. The sudden sign change
164: of a cutting force in a function of relative velocity $v_f$
165: is due to a friction
166: phenomenon between the tool and a chip.
167:
168: Our model (Eqs. 1-3) with one degree-of-freedom is a serious simplification of
169: a physical situation.
170: However our aim is not the comprehensive description of a cutting process.
171: Here we want to concentrate on particular aspects of it. In spite of
172: simplicity of the examined model still allows the chatter vibrations to be generated due to
173: the nonlinearities in of the cutting force $g_y(h,v_f)$ as it was shown in paper \cite{15}.
174: In our model chatter is generated by
175: a combination of the friction phenomenon between the tool and chips, and
176: the impact of a tool after loosing its contact with a workpiece.
177: Warmi\'nski and others \cite{15} examined the second pass of the orthogonal cutting process
178: by using similar
179: model. The results
180: obtained there have indicated that such model can lead to periodic, quasi-periodic as well as chaotic
181: vibrations due to the initial harmonic modulation of the machined surface.
182:
183: For a numerical calculation purpose we have written the equations (Eqs.
184: 1-4) in
185: discrete way introducing the constant
186: time step $\tau$:
187: \begin{eqnarray}
188: t_{r+1} &=& t_r + \tau, \nonumber \\
189: y_{r+1} &=& y_r +v_r \tau, \\
190: v_{r+1} &= &v_r + (-2 \tilde n v_r - p^2 y_r + \frac{K}{m} g_{yr}) \tau,
191: \nonumber
192: \end{eqnarray}
193: where $t_r$ is a sampling discrete time
194: after $r$ time steps.
195: The function $g_{yr}$ should be expressed:
196: \begin{eqnarray}
197: g_{yr} &=& g_y (h_r, v_{fr}), \nonumber \\
198: h_r &=& h_0+y_s -y_r, \\
199: v_{fr} &=& 1- \frac{v_r}{v_0}, \nonumber
200: \end{eqnarray}
201: where $r$ and $s$ are natural numbers.
202: The time difference between $y_r$ and $y_s$ coordinates; $\Delta t =(r-s)
203: \tau $
204: relates to the time of workpiece revolution (Eq. 4). The system
205: parameters
206: obtained from the experiment are
207: following: $p=785$ rad/s, $m=12.1$kg, $K=620$ N, $h_0=1.5 \times 10^{-3}$
208: m,
209: $f=0.1~\times 10^{-3}$ m/rev, $2 \tilde n =190~ $1/s, $v_0=0.1$ m/s $\kappa=
210: 70^o$ and $c_2=0.5$,
211: $c_3=1.55$ are cutting process constants derived from \cite{7,15}.
212:
213:
214: \section{INFLUENCE OF NOISE}
215:
216: Most of real dynamical processes are disturbed by the random signal. In
217: case
218: of cutting process they come through the
219: roughness of the initial surface (Fig. 1). Other sources of such
220: disturbances can be found in a spontaneous breaking of chips and
221: the couplings of the tool and the workpiece to other
222: dynamic parts of the experimental standing.
223:
224: To describe the stochastic system we introduce to the model a random
225: component by an additive white noise of Gaussian distribution \cite{10,11,12,13,14}.
226: Usually stochastic dynamic systems are investigated by using the Fokker-Planck equation
227: \cite{8,9,10,11}.
228: One dimension version of it reads:
229: \begin{equation}
230: \frac{\partial}{\partial t} P(y,t) = \frac{\partial}{\partial y} \left[ \nu (y)
231: P(y,t)\right] + D \frac{\partial^2}{\partial y^2} P(y,t),
232: \end{equation}
233: where $D$ denotes the diffusion coefficient, $\nu (x)$ is, in general, the non-linear drift term
234: (driving
235: force) and $P( x,t)$ the probability distribution function.
236: As solving Fokker-Plank equation meets some considerable difficulties \cite{10,11}
237: we have transformed it into the corresponding Langevin equation:
238: \begin{equation}
239: \dot y = z(y) + g \Gamma (t),
240: \end{equation}
241: where $g \Gamma (t)$ is a Gaussian distributed random 'force' with the strength $g$ and
242: $\Gamma (t)$ is assumed to satisfy:
243: \begin{eqnarray}
244: < \Gamma(t) > & = & 0, \\
245: < \Gamma(t), \Gamma(t') > & = & 2 \delta( t-t'), \nonumber
246: \end{eqnarray}
247: where the brackets denote an average over the probability distribution function.
248: Starting with the definitions of the drift $\nu$ and diffusion $D$ coefficients obtained by
249: the Kramers-Moyal expansions in the derivation of Fokker-Plank equation from a
250: Chapman-Kolmogorov equation \cite{17}, we can find the relation between $\nu(y)$ and $D$ of the
251: Fokker-Plank equation (Eq. 7) with $z(y)$ and $g$ of the Langevin equation (Eq. 8). Thus
252: the
253: Langevin equation can be finally expressed by drift and diffusion terms of the
254: initial Fokker-Plank
255: equation as:
256: \begin{equation}
257: \dot y = \nu(y) + \sqrt{D} \hat \Gamma (t)
258: \end{equation}
259: For the actual numerical calculation we have discretized form:
260: \begin{equation}
261: y(t+\tau)-y(t)= \int_{t}^{t+\tau} {\rm d} t' v(t') + \sqrt{D}
262: \int_{t}^{t+\tau} {\rm d} t' \Gamma (t') \approx \nu (t) \tau +
263: \sqrt{D} \hat \Gamma (t),
264: \end{equation}
265: where
266: \begin{equation}
267: \hat \Gamma (t) =\int_{t}^{t+\tau} {\rm d} t' \Gamma (t')
268: \end{equation}
269: is a superposition of Gaussian distributed random numbers which again are
270: of a Gaussian form. Namely:
271: \begin{equation}
272: \hat \Gamma (t) = a \omega (t).
273: \end{equation}
274: In our case the average value of $\omega$ has been chosen as $<\omega>
275: =0$ while
276: its variance as
277: $<\omega^2>=2$, respectively. From the integration of the Gaussian (Eq. 12) the
278: coefficient
279: $a$ (Eq. 13) depends on the time integration step
280: $\tau$ via $a = \sqrt{\tau}$.
281: Equation 8 can be, in general,
282: solved by higher order
283: algorithms like Runge-Kutta one \cite{18}. However here, for simplicity,
284: we
285: have
286: limited our discussion to the simplest algorithm of Euler type. Thus, the final
287: form of
288: the Langevin equation in the lowest order perturbation, suitable for
289: numerical
290: integration is given by the following
291: expression:
292: \begin{equation}
293: y_{r+1}=y_r + \nu (y_r)\tau + \sqrt{D} \sqrt{\tau} \omega (t_r).
294: \end{equation}
295:
296: Here we analyzed the one-dimensional version of the Langevin and Fokker-Plank equations.
297: The similar discussion on the $m$-dimensional stochastic Langevin equation and its relation
298: to
299: the corresponding Fokker-Plank equation can be
300: found in
301: \cite{14}.
302:
303:
304: In our model, Eq. 9 for $y_r$ has to be supplemented by the rest of
305: equations (Eqs. 5,6) for discrete time $t_r$ and velocity $v_r$
306: which is now substituted by the corresponding drift term $\nu_r$:
307:
308: \begin{equation}
309: \nu_{r+1} = \nu_r + (-2 \tilde n v_r - p^2 y_r + \frac{K}{m} g_{yr}) \tau.
310: \end{equation}
311:
312:
313:
314: Using the above procedure (Eqs. 4,5 and Eq. 14-15) we have done the
315: simulations
316: for
317: a constant time step $\tau =0.741 \times 10^{-4}$ s,
318: corresponding to the workpiece revolution time $\Delta t=0.741 \times 10^{-1}$ s
319: and a number of diffusion constants values $D$. Figures 4 a--d show the
320: time
321: histories of $y$ of initial 3 seconds
322: of cutting work for experimentally identified system parameters (Sec. 2).
323: For a deterministic system ($D=0$, Fig. 4a)
324: we observe the stable cutting
325: process with no vibrations. Figures 4b, c relate to cutting process in
326: presence of noise. The diffusion constant values
327: in these figures are $D=10^{-5}$ and $D=10^{-4}$ respectively. One can see
328: easily the presence
329: of small vibrations (Fig. 4b), which
330: are growing with increasing of the noise level (Fig. 4c). Obviously, such vibrations have a
331: significant effect on the quality of
332: a workpiece surface. It is shown in
333: Fig. 5, where the error shape of a surface is plotted as a function of
334: a workpiece rotation angle after 3 s of cutting
335: work.
336:
337:
338:
339:
340:
341:
342:
343:
344: Modulation of shape caused by random disturbances depends on a noise
345: level.
346: Both:
347: input and output random signals can be
348: easily measured by standard deviations. For various $D=10^{-6}$, $10^{-5}$
349: and $10^{-4}$
350: we have got the following values of
351: standard deviations: $\sigma= 3.79 \times 10^{-
352: 7} $ m, $1.21 \times 10^{-6}$ m and $4.16 \times 10^{-6}$ m. In
353: Figs. 6a and b we have compared
354: the input and output random signals for
355: one of the above cases ($D=10^{-5}$). Figure 6a shows the distribution of
356: Gaussian disturbances of input noise $\omega$ (Eqs.
357: 8,9) while Fig. 6b corresponds to the errors
358: of workpiece shape after cutting. In Fig. 6b deviation
359: from the normal probability
360: distribution is caused by nonlinear dynamics of the cutting process. To
361: quantify the system answer we have used
362: standard deviations of the output signal
363: $\sigma$ as a function of diffusion constant $D$. It is plotted in
364: the logarithmic
365: scale in Fig. 7. We have checked that
366: $\sigma (D)$ can be scaled as a square root as far as noise level is low
367: while
368: for stronger noise its effect on the
369: fluctuations of $y$ is stronger.
370:
371:
372:
373: \section{SUMMARY AND CONCLUSIONS}
374: We have considered vibrations of a tool-workpiece system in a straight turning
375: process induced by
376: random disturbances and their effect on a product surface.
377: Using a single degree-of-freedom model we have focused on the
378: combined effects of the friction nonlinearities and tool-workpiece contact loss.
379: We have
380: noticed that for large
381: enough level of noise, the tool and a workpiece start to
382: vibrate due to random forcing. Such excitation can interact with a complex
383: dynamics of the system leading
384: to process non-stability and,
385: in the end, to much worse final quality of the machined product.
386: In case of a relatively small level of noise (a small value of the diffusion constant $D$) the surface
387: shape
388: error scales as a square root of $D$. For a higher value of a noise level the shape error is proportional to $D$.
389: Clearly, in a straight turning the initial surface roughness influence the quality of a final product.
390: This is the principal result of our paper, which lead to a conclusion that
391: one has to prepare workpiece which initial surface satisfies the appropriate criteria.
392:
393:
394:
395: \begin{ack}
396: The work has been partially supported by Polish State Committee for
397: Scientific
398: Research (KBN)
399: under the grant No. 126/E-361/SPUB/COST/T-7/DZ 42/99.
400: We would like to thank the organizers of $3^{rd}$ International Symposium
401: "Investigation of Nonlinear Dynamic Effects in Production Systems" 26-27
402: September 2000 in Cottbus (Germany)
403: for giving two of us (J.L. and R.R.) opportunity to present this work.
404: We would like to thank the unknown referee for
405: valuable comments and Dr W. Przystupa for helpful discussions.
406: \end{ack}
407: \begin{thebibliography}{99}%\s1
408: \bibitem{1}
409: I. G{\sc RABEC} {1986} {\it Phys.
410: Lett.}
411: {\bf A 117}, {384--386}, Chaos Generated by the Cutting Process.
412: \bibitem{2}
413: I. G{\sc RABEC} {1988}
414: {\it J. Mach.
415: Tools Manufact.} {\bf 28} {19--32}, Chaotic Dynamics of the Cutting
416: Process.
417: \bibitem{3}
418: J. G{\sc RADISEK}, E. G{\sc OVEKAR}, I. G{\sc RABEC} {1996}
419: {\it Int. J. Mach. Tools Manufact.} {\bf 36} {1161--1172},
420: A Chaotic Cutting
421: Process and
422: Determining Optimal Cutting Parameter
423: Using Neural Networks.
424: \bibitem{4}
425: I.N. T{\sc ANSEL}, C. E{\sc RKAL}, T. K{\sc ERAMIDAS} {1992}
426: {\it Int. J. Mach.
427: Tools Manufact} {\bf 32} {811--827}, The Chaotic
428: Characteristic of
429: Three Dimensional Cutting.
430: \bibitem{5} E. M{\sc ARUI}, S. K{\sc ATO}, M. H{\sc HASHIMOTO}, T. Y{\sc
431: AMADA} {1988} {\it ASME Journal of Engineering for Industry} {\bf 110}
432: {242--247}, The Mechanism of Chatter Vibrations in a Spindle-Workpiece
433: System: Part 2 - Characteristics of Dynamic Cutting Force and Vibration
434: Energy.
435: \bibitem{6} G. L{\sc ITAK}, J. W{\sc ARMI\'NSKI}, J. L{\sc IPSKI}
436: {1997} in {\it Proceedings of 4th Conference on Dynamical Systems - Theory
437: and
438: Applications L\'od\'z 8-9 December 1997}, Eds J. Awrejcewicz, J. Grabski
439: and J. Mrozowski, L\'od\'z, {193-197},
440: Self-Excited Vibrations in Cutting Process.
441: \bibitem{7} M. W{\sc IERCIGROCH} {1997} {\it ASME Journal of Vibration
442: and Acoustics} {\bf 119}, {468-475} Chaotic Vibrations of a Simple Model
443: of the Machine Tool - Cutting System.
444: \bibitem{8}
445: M. W{\sc IERCIGROCH}, A. H-D. C{\sc HENG} {1997}
446: {\it Solitons \& Fractals} {\bf 8} {715-726}, Chaotic and Stochastic
447: Dynamics
448: of Orthogonal Metal Cutting.
449: \bibitem{9}
450: T. K{\sc APITANIAK} {1990} {\it Chaos in Systems with Noise}, 2nd
451: revised edn.,
452: Singapore, World Scientific.
453: \bibitem{10}
454: H. R{\sc ISKEN} {1984} {\it Fokker-Planck Equation: Methods of Solution
455: and
456: Applications}, Berlin, Springer Verlag.
457: \bibitem{11}
458: W. P{\sc RZYSTUPA}, G. L{\sc ITAK} {1998} in
459: {\it Zeszyty Naukowe Katedry
460: Mechaniki Stosowanej, Gliwice, Proceedengs of XXXVII Sympozjon Modelling
461: in
462: Mechanics
463: Wisla February 1998}, Ed. E Opoka,
464: Politechnika \'Sl\c{a}ska, Gliwice 1998) {304--310}, Cutting Process in
465: Presence of Noise.
466: \bibitem{12}
467: J. W{\sc ARMI\'NSKI}, G. L{\sc ITAK}, K. S{\sc ZABELSKI} 2000
468: in {\it Applied Nonlinear Dynamics and Chaos of Mechanical Systems with
469: Discontinuities} Eds. M. Wiercogroch and B. De
470: Kraker, {\it Series on Nonlinear Science Series} {\bf A} Vol. {\bf 28}
471: (World Scientific
472: Singapore) {197-200} Dynamic Phenomena in Gear Boxes. Influence of Noise.
473: \bibitem{13}
474: J. G{\sc RADISEK}, I. G{\sc RABEC} {1999}
475: in Proceedings of COST P4 Meeting, Wroclaw, 15th October 1999,
476: Stochastic and Nonlinear
477: Modelling of a
478: Cutting Process.
479: \bibitem{14}
480: J. G{\sc RADISEK}, S. S{\sc IEGERT}, R. F{\sc RIDRICH AND}
481: I. G{\sc RABEC} {2000} {\it Phys. Rev.} {\bf E 62} 3146-3155
482: Analysis of Time Series from Stochastic Processes.
483: \bibitem{15}
484: J. W{\sc ARMI\'NSKI}, G. L{\sc ITAK}, J. L{\sc IPSKI}, M. W{\sc
485: IERCIGROCH},
486: M.P. C{\sc ARTMELL} 2000
487: in: {\it Solid Mechanics and Its Applications} Vol. {\bf 73}, Eds. E.
488: Lavendelis
489: and M.
490: Zakrzhevsky
491: (Kluver Academic Publisher) {275--283},
492: Vibrations in Regenerativ Cutting Process Synthesis of Nonlinear
493: Dynamical
494: Systems.
495: \bibitem{16}
496: G. S{\sc TEPAN}, T. K{\sc ALMAR}-N{\sc AGI} {1997} in
497: {\it Proceedings of 16th ASME Biennial Conference on Mechanical Vibrations and Noise,
498: ASME Design and Technical Conferences 1997, September 14-17, 1997, Sacramento,
499: California}, 1--11, Nonlinear Regenerative
500: Machine
501: Tool Vibrations.
502: \bibitem{17} P. F{\sc R\"{O}BRICH} and S.Y. X{\sc U} {1988} {\it Nuclear Physics} {\bf A477}, {143--161}
503: The Treatment of Heavy-Ion Collisions by Langevin Equations.
504: \bibitem{18} R.L. H{\sc ONEYCUTT} {1991} {\it Phys. Rev.} {\bf A 45},
505: {600--604}, Stochastic Runge--Kutta Algorithms. I. White Noise.
506: \end{thebibliography}
507:
508: \newpage
509:
510: {\center FIGURE CAPTIONS}
511:
512: \noindent
513: Figure 1. Experimental standing with a cast iron workpiece. \\
514: Figure 2. Physical model of a straight turning process. \\
515: Figure 3. Nonlinear function $g_y(h,v_f)$ versus cutting relative
516: velocity. \\
517: Figure 4. Time histories of $y$ for various values of diffusion constants
518: (a) $D=0$, (b) $D=10^{-5}$, (c) $D=10^{-4}$. \\
519: Figure 5. Shape error as a function of workpiece rotation angle after $3
520: s$ of cutting. \\
521: Figure 6. Distribution probabilities of random input (a) and output $y$
522: (b) signals of the model for $D=10^{-5}$. \\
523: Figure 7. Standard deviation $\sigma$ as a function of a diffusion
524: constant
525: $D$.
526:
527: \end{document}
528:
529: