1: \documentstyle[12pt]{article}
2:
3: \oddsidemargin 5mm
4: \evensidemargin 5mm
5: \textwidth=16.5cm
6: \topmargin -5mm
7: \textheight=22.5cm
8:
9: \begin{document}
10:
11: \begin{center}
12: {\large {\bf SELF-SIMILAR EXTRAPOLATION FOR THE LAW OF
13: ACOUSTIC EMISSION BEFORE FAILURE OF HETEROGENEOUS MATERIALS} \\ [3mm]
14:
15: A. Moura$^1$ and V.I. Yukalov$^2$} \\ [2mm]
16: {\it $^1$GEMPPM, UMR CNRS 5510 (Bat. B. Pascal)\\
17: Institut National des Sciences Appliquees de Lyon \\
18: 20, av. A. Einstein, 69 621 Villeurbanne, France \\
19: e-mail: andre.moura@insa-lyon.fr \\ [3mm]
20:
21: $^2$Bogolubov Laboratory of Theoretical Physics \\
22: Joint Institute for Nuclear Research, Dubna 141980, Russia \\
23: e-mail: yukalov@thsun1.jinr.ru}
24: \end{center}
25:
26: \vskip 2cm
27:
28: \begin{abstract}
29:
30: Acoustic emission before the failure of heterogeneous materials
31: is studied as a function of applied hydrostatic pressure. A formula for
32: the energy release is suggested, which is valid in the whole diapason of
33: pressures, from zero to the critical pressure of rupture. This formula is
34: obtained by employing the extrapolation technique of the self-similar
35: approximation theory. The result is fitted to experiment in order to
36: demonstrate the correct general behaviour of the obtained expression for
37: the energy release.
38: \end{abstract}
39:
40: \section{Introduction}
41:
42: There exists a widespread understanding in scientific community that the
43: moment of rupture in heterogeneous materials is somewhat similar to a
44: critical point. The fracturing process is then a kind of a second-order
45: phase transition [2,4]. This underlies the intuitive assumption which compares
46: global failure in disordered materials at mesoscopic scale with percolation
47: at microscopic scale. So, as percolation, global failure is assumed to be
48: a critical phenomenon. The fracturing process is then classified as a
49: self-organised random irreversible process. Extensive numerical simulations
50: and experimental measurements have confirmed this conclusion. This concerns
51: the global failure of materials as well as earthquakes [9], since an
52: earthquake can also be considered as a global failure of a large material
53: mass. The classical power law, defining the energy release rate, has been
54: verified in the critical region, close to the time of either a large
55: earthquake or a global materials failure. The related critical exponent
56: is, in general, a complex number, which is due to taking into account
57: long-range elastic interactions transported by stress fields around defects
58: and cracks. The imaginary part of the exponent gives rise to the so-called
59: log-periodic corrections which have been identified quite early in
60: renormalization group solutions for critical phase transitions [7,10].
61: Experimental results of acoustic emission measurements and seismograms
62: unambiguously revealed the existence of oscillatory corrections before
63: materials fractures and earthquakes, respectively [1,10]. If one would
64: know the general law describing this acoustic emission, one would be able
65: to make an early prediction of the global failure in heterogeneous materials.
66: It is a principal, not yet solved problem, to find such a general law,
67: which would be valid not solely in the asymptotic vicinity of the critical
68: point, but in the whole interval of pressures or times, both near the critical
69: point of failure as well as far from it. And this is the main aim of the
70: present paper to suggest a mathematically grounded derivation for such
71: a law and to confront it to available experimental data [5]. For this purpose,
72: we employ the recent progress in the self-similar approximation theory [12--18]
73: which has been successfully applied to the theory of critical phenomena and
74: to time series forecasting [19--21]. We would like to stress here that this
75: theory is general and can be used for any given series of experimental data,
76: whether this concerns phase transitions, financial markets, or any other
77: series.
78:
79: \section{Modelling}
80:
81: Let us be interested in the behaviour of the energy release $E(p)$
82: as a function of pressure $p$, considered between $p\approx 0$, i.e. at the
83: initial stage when the very early damages occur, and up to the point $p=p_c$
84: of global failure. Note that $p$ refers to fracturing under applied spatially
85: uniform loading, as in experiments where water pressure inside a spherical tank
86: increases up to the material failure. Experimentally, the applied pressure is
87: usually increased linearly with time. The behaviour of $E(p)$ at the vicinity
88: of the critical point of rupture is
89: \begin{equation}
90: \label{1}
91: E(p) \approx E(p_c) + A(p_c - p)^\alpha + B(p_c-p)^\alpha\cos[\omega\ln
92: (p_c-p)+\varphi] \; ,
93: \end{equation}
94: as $p\rightarrow p_c$, where $A,B,\alpha,\omega$ and $\varphi$ are parameters
95: ($\alpha$ and $\omega$ are the real and imaginary parts, respectively,
96: of the so-called critical complex exponent). The problem is how to find the
97: behaviour of $E(p)$ in the whole range $0\leq p \leq p_c$? Here we consider
98: the case of one variable, pressure. A more general loading, involving not one
99: variable, like $p$, but several variables can also be treated by the theory.
100: However, before being involved in such generalisations, we would like to show
101: on a simpler example how to accomplish the extrapolation. Our main aim in
102: this letter is to demonstrate the general possibility of realising such an
103: accurate extrapolation for the energy release.
104:
105: Let us transform Eq. (1) by introducing the dimensionless variable
106: $x\equiv(p_c-p)/p_c$, $(0\leq x\leq 1)$ and the reduced energy release
107: $f(x)\equiv E(p)/E(p_c)$ which is a dimensionless function. Then, expansion
108: (1) takes the form
109: \begin{equation}
110: \label{2}
111: f(x) \approx 1 + \tilde a(x)\; x^\alpha \; , \qquad
112: \tilde a(x) \equiv \lambda[1+\mu\cos(\omega\ln x +\beta)]
113: \end{equation}
114: whose parameters $\beta$, $\lambda$, and $\mu$ can be easily expressed through
115: $p_c$, $E(p_c)$, $A$, $B$, $\omega$, and $\varphi$. Note that function
116: $\tilde a(x)$ cannot be expanded in powers of $x$ near zero. Therefore,
117: expansion (2) can be treated as a generalised asymptotic series, as defined
118: by Poincare [8], with $\tilde a(x)$ considered as a coefficient. Employing
119: the self-similar approximation theory [12--21], the asymptotic series (2)
120: can be extrapolated to
121: \begin{equation}
122: \label{3}
123: f^*(x)=\exp[c(x)\; x^\alpha] \; , \qquad c(x) = [1 +\mu\cos(\omega\ln x +
124: \beta)]\; \tau \; ,
125: \end{equation}
126: where $\tau$ has to be defined from an optimization or boundary condition.
127: As is clear, the boundary condition $E(0)\approx 0$ is not convenient here,
128: leading to large errors in defining $\tau$ because of the physical impossibility
129: to determine the very early precursors with a high accuracy. A more judicious
130: choice is to consider a global physical quantity. In the same way as treating
131: $E(p)$ as the integrated energy release rate, we may define $F(p)$ as the integral
132: energy release $\int_0^p E(q)dq$. Given $p_0$, the parameter $\tau$ then can be
133: defined from the sum optimization rule
134: \begin{equation}
135: \label{4}
136: \int_{y_0=\frac{p_c-p_0}{p_c}}^1 \; f^*(x)\; dx = \frac{F(p_0)}{p_cE(p_c)} \qquad
137: (y_0\rightarrow 0, \; p_0\rightarrow p_c-0) \; .
138: \end{equation}
139: In general, the lower limit in the integral (4) can pertain to the whole interval
140: [0,1]. In reality, the trustful experimental data are available only below $p_c$.
141: One has to choose the data at the highest available pressure, so that the lower
142: limit in the sum rule (4) be sufficiently small.
143:
144: In this way, the self-similar approximant (3) becomes
145: \begin{equation}
146: \label{5}
147: f^*(x) =\exp\left\{ \left [ 1 +\mu\cos(\omega\ln x +\beta)\right ] \;
148: \tau x^\alpha \right \}
149: \end{equation}
150: with $\tau$ given by the sum rule (4). Note that this formula contains the same
151: number of parameters as the initial formula (1). Returning, with the help
152: of relations (1) and (2), to the energy release, we come to the formula
153: \begin{equation}
154: \label{6}
155: E^*(p) = E(p_c)\; f^*(x) \; ,
156: \end{equation}
157: which extrapolates $E(p)$ for the whole interval $0\leq p\leq p_c$. It is worth
158: stressing that the law (6) is obtained as a self-similar extrapolation of the
159: acoustic emission signals observed at the vicinity of the rupture. Another
160: possibility would be to base this kind of extrapolation, starting from the signals
161: existing at the initial stage of the process, being yet very far from the rupture.
162: This latter way was used by Gluzman et al. [3], who started with a given
163: polynomial expansion at the early time of acoustic emissions, treating the power
164: law behaviour close to rupture as a boundary condition. To our mind, this opposite
165: approach has two weak points: First of all, the early acoustical precursors, as
166: is known, are very small, being embodied in the acoustical noise, because of
167: which they can be hardly measured with a good accuracy. Second, it looks
168: difficult or even impossible to extract information on log-periodic corrections
169: from the early acoustic signals. Another phenomenological expression for the
170: energy release in the noncritical region has been suggested by Sornette and
171: Andersen [10] whose arguments were based on the existence of a scaling of the
172: macroscopic elastic modulus and the elastic energy release rate in the thermal
173: fuse model [6]. The following phenomenological form has been proposed [5]
174: \begin{equation}
175: \label{7}
176: E(p) \approx E(p_c) A\left ( {\rm tanh}\;\frac{p_c-p}{\tau_0}\right )^\alpha +
177: B\left ( {\rm tanh}\;\frac{p_c-p}{\tau_0}\right )^\alpha\;
178: \cos\left\{ \omega\left ( {\rm tanh}\;\frac{p_c-p}{\tau_o}\right ) +\varphi
179: \right \} \; ,
180: \end{equation}
181: which is a pure power law, like Eq. (1), in the critical region close to
182: rupture and exponentially relaxes in the noncritical region far from rupture,
183: where only a few damages occur. However, the suggested form (7) possesses an
184: unphysical behaviour, being negative in a wide range of its variable. At the
185: same time, our self-similar formula (5) yields the energy release (6) that
186: is always positive. In Fig. 1, we compare the behaviour of $E(p)$ , in the
187: dimensionless form, for both Eqs. (6) and (7), with the parameters chosen
188: so that to have similar variations when approaching the critical point. In
189: this figure, one can clearly see that Eq. (7) possesses unphysical negative
190: values for the major part of the interval [0,1], while Eq. (6) is everywhere
191: positive.
192:
193: \section{Comparison with experimental data}
194:
195: In order to fit our self-similar formula to experimental results of acoustic
196: emission measurements, we refer to the most accurate data, available nowadays,
197: that is those of Anifrani's team [1] at Aerospatiale-Matra Inc., which are
198: reported by Johansen and Sornette [5]. In these measurements, the acoustic
199: emission has been recorded during the pressurisation of spherical tanks of
200: kevlar or carbon fibres pre-impregnated in a resin matrix wrapped up around
201: a thin metallic liner (steel or titanium) fabricated and instrumented by
202: Aerospatiale-Matra Inc. It has been found that the seven acoustic emission
203: recordings of seven pressure tanks, which were brought to rupture, exhibit
204: a clear acceleration in agreement with a power law divergence as expected by
205: the critical point theory. A strong evidence of oscillatory corrections,
206: forming the intermittent succession of bursts of acoustic emission, when
207: approaching the rupture, has been clearly identified.
208:
209: In Fig. 2, we present the energy release rate, which is the derivative of our
210: formula (6) for the cumulative energy release. The result is obained by fitting
211: to the experimental curve reported at the right bottom in figure 1 of the paper
212: by Johansen and Sornette (2000). These experimental data explicitly show an
213: oscillatory behaviour of the energy release rate before the rupture. The best
214: fit for the five parameters $\alpha$, $\beta$, $\mu$, $\tau$, and $\omega$
215: have been obtained for 0.7, 1.9, -0.045, 7.5 and 15, respectively, under the
216: given pressure of rupture at 673 bars. The parameter $\tau$ satisfies Eq. (4).
217: The upper pressure $p_0\approx 665$ bars, measured before the rupture, corresponds
218: to the value $y_0\approx 0.0119$, which is sufficiently small, in agreement with
219: condition (4).
220:
221: \section{Conclusion}
222:
223: The parameters can change for different materials as well as for different
224: experimental setups. However, the three key physical parameters: $\alpha$,
225: $\omega$, and $\tau$ should be universal, not essentially depending on the
226: details of the external stress in the critical region. To prove this assumption,
227: it is necessary to accomplish a set of experiments for appropriate materials.
228: Selecting particular materials, we should keep in mind that: (i) the width of
229: the critical region is linearly proportional to the strength of disorder,
230: (ii) long-range elastic interactions are responsible for log-periodic corrections,
231: (iii) the more brittle is the fracturing process, the stronger are the elastic
232: interactions and stronger the energy release fluctuates, (iv) the fluctuations
233: in the energy release are also stronger when a high value of a material
234: characteristic is alternated with a low value of that characteristic. Thus,
235: elastic-porous materials are expected to be the best choice.
236:
237: \newpage
238:
239: \begin{thebibliography}{99}
240: \bibitem{1}
241: Anifrani J. C., Le Floch C., Sornette D. and Souillard B.,
242: Universal log-periodic correction group scaling for rupture stress
243: prediction from acoustic emission. J. Phys. I France 5 (1995) 631-638.
244: \bibitem{2}
245: Broberg B., Cracks and Fractures. Cambridge, London (1999).
246: \bibitem{3}
247: Gluzman S., Andersen J. V. and Sornette D.,
248: Functional renormalization prediction of rupture. In A. Levshin, G. Molchan
249: and B. Naimark (eds.) Computational Seismology. Moscow, GEOS (2001).
250: \bibitem{4}
251: Herrmann H. J. and Roux S.,
252: Statistical Models for the Fracture of Disordered Media.
253: North-Holland, Amsterdam (1990).
254: \bibitem{5}
255: Johansen A. and Sornette D.,
256: Critical ruptures. Eur. Phys. J. B 18 (2000) 163-181.
257: \bibitem{6}
258: Lamagnere L., Carmona F. and Sornette D.,
259: Experimental realization of critical thermal fuse rupture.
260: Phys. Rev. Lett. 77 (1996) 2738-41.
261: \bibitem{7}
262: Nauenberg M. et al., Scaling representation for critical phenomena.
263: J. Phys. A 8 (1975) 925-928.
264: \bibitem{8}
265: Poincare H., New Methods of Celestial Mechanics. Am. Inst. Phys.,
266: New York (1993).
267: \bibitem{9}
268: Sahimi M. and Arbabi S., Scaling laws for fracture of heterogeneous materials
269: and rocks. Phys. Rev. Lett. 77 (1996) 3689-3692.
270: \bibitem{10}
271: Sornette D. and Andersen J. V., Scaling with respect to disorder.
272: Eur. Phys. J. B 1 (1998) 353-357.
273: \bibitem{11}
274: Sornette D., Discrete scale invariance and complex dimensions. Phys. Rep.
275: 297 (1998) 239-270.
276: \bibitem{12}
277: Yukalov V.I., Statistical mechanics of strongly nonideal systems.
278: Phys. Rev. A 42 (1990) 3324-3334.
279: \bibitem{13}
280: Yukalov V.I., Method of self-similar approximations. J. Math. Phys.
281: 32 (1991) 1235-1239.
282: \bibitem{14}
283: Yukalov V.I., Stability conditions for method of self-similar approximations.
284: J. Math. Phys. 33 (1992) 3994-4001.
285: \bibitem{15}
286: Yukalov V.I. and Gluzman S., Critical indices as limits of control functions.
287: Phys. Rev. Lett. 79 (1997) 333-336.
288: \bibitem{16}
289: Yukalov V. I. and Gluzman S., Self-similar bootstrap of divergent series.
290: Phys. Rev. E 55 (1997) 6552-6565.
291: \bibitem{17}
292: Yukalov V.I. and Gluzman S., Self-similar exponential approximants.
293: Phys. Rev E 58 (1998) 1359-1382.
294: \bibitem{18}
295: Yukalov V.I., Yukalova E.P. and Gluzman S.,
296: Self-similar interpolation in quantum mechanics, Phys. Rev. A 58 (1998) 96-115.
297: \bibitem{19}
298: Yukalov V. I. and Gluzman S.,
299: Weighted fixed points in self-similar analysis of time series.
300: Int. J. Mod. Phys. B 13 (1999) 1463-1476.
301: \bibitem{20}
302: Yukalov V. I.,
303: Self-similar extrapolation of asymptotic series and forecasting for time series.
304: Mod. Phys. Lett. B 14 (2000) 791-800.
305: \bibitem{21}
306: Yukalov V.I.,
307: Self-similar approach to market analysis. Eur. Phys. J. B 20 (2001) 609-617.
308: \end{thebibliography}
309:
310: \newpage
311:
312: \begin{center}
313: {\large{\bf Figure Captions}}
314: \end{center}
315:
316: \vskip 2cm
317:
318: {\bf Fig. 1}. General form of the dimensionless energy release corresponding to
319: Eq. (6), with $\alpha=0.7$, $\omega=6.9$ and $\tau=6.6$ (continuous line) and
320: to Eq. (7), with $\alpha=0.7$, $\omega =8$, and $\tau_0=0.8$ (broken line).
321:
322: \vskip 2cm
323:
324: {\bf Fig. 2}. Dimensionless energy release rate from Eq. (5), fitted to the
325: experimental data reported by Johansen and Sornette (2000).
326:
327:
328: \end{document}
329: