Acoustic Emission Study of Micro- and Macro-fracture of Large Rock Specimens Xiang-chu YIN 1,2, Huai-zhong YU 1, Ke-yin PENG 2, Victor Kukshenko 3, Zhaoyong XU5, Qi LI 4, Meng-fen XIA 1,6, Min LI 1, and Surguei Elizarov 7 1, State Key Laboratory of Nonlinear Mechanics, CAS 2, Center for Analysis and Prediction, China Seismological Bureau 3, Ioffe Physical Technique Institute, Russian Academy of Sciences 4, Ibaraki University, Japan 5, Yunnan Province Seismological Bureau, CSB 6, Peking University, China 7, Interunis Ltd, Moscow, Russia

Introduction There are striking resemblances between AE (acoustic emissions) and earthquakes. Consequently to study the AE during the process of micro- and macro-fracture in rock will help us to understand the nature of earthquake..

In the first half of this year, we conducted a series of experiments with rectangular prisms of three kinds of rocks (Dali marble, Wuding gneiss and Wuding sand- stone). Three large specimens of each kind of rock have been conducted. The geometry of the large specimen is 105X40X15 cm 3 so the large size of the specimen reaches to more than 1 meter

z q x y f The specimen is loaded in two directions: the axial stress σ 1 and lateral stress σ 2 Another principal stress σ 3 is zero so that: σ 1 ≠σ 2 ≠σ 3.

Loading history There are two kinds of loading history : monotonously loading and cycling loading.

Experiment results At first I present the results of cycling loading. The AE signals are recorded continuously with 《 A-line 32D---AE system 》 made by A.F Ioffe Physical Technical Institute, Russian Academy of Sciences and Interunis Ltd.

The 《 A-line 32D---AE system 》 is a 32 channels AE system. Each channel consists of an AE sensor, a preamplifier and an AECB board(Acoustic Emission Channel Board). AE sensor pick up the stress wave from the specimen and convert it into an electronic signal which is then amplified by a preamplifier and converted into a digital data stream in a AESB. AE features such as arrival time, rise-time, duration, pick amplitude, energy and counts are extracted by a FPGA (Field Programmable Gate Array). In parallel to the feature extraction, the complete waveform can also be stored (in an optional OSC recorder module) and recorded.

level a level b level c level d

Gneiss 3(a)

Gneiss 3(c)

Gneiss 3(d)

Gneiss 3

Small gneiss 2

From these figures we can see the validity of Kaiser effect for rock is seriously questioned. In our cycling expe- riments, all the loading peaks are the same, but for the second and ensuing cycles their AE activity are still active, even though the activity decrease gradually with the cycle number. Every peak of load correspond a peak of AE. The peak of AE lags behind the corresponding peak of load about a minute in order.

AE location distribution In the mean time the AE events can be located in real time and can be shown on the screen and be recorded.

LURR The Load-Unload Response Ratio (LURR) is defined as (1) where X + and X - are the response rates during loading and unloading according to some measure.

The idea that motivated the LURR earthquake prediction approach is that when a system is stable, its response to loading is nearly the same as its response to unloading so LURR ~ 1, whereas when the system is approaching an unstable state, the response to loading and unloading becomes quite different and LURR >1.

High LURR values (larger than unity) indicate that a region is prepared for a large earthquake. In previous years, a series of successful intermediate-term predictions have been made for strong earthquakes in China and other countries using the LURR (YIN and YIN, 1991; YIN, 1993; YIN et al., 1994; YIN et al., 1995; YIN et al., 1996; YIN et al., 2000). Usually the released seismic energy is adopted as the “response” and then the LURR is defined as:

where E denotes seismic energy, the sign “+” means loading and “–” means unloading, m=0 or 1/3 or1/2 or 2/3 or 1. When m=1, E m is exactly the energy itself; m=1/2, E m denotes the Benioff strain; m=1/3, 2/3, E m represents the linear scale and area scale of the focal zone respectively; m=0, Y is equal to N+/N–, and N+ and N– denote the number of earthquake occurred during the loading and unloading duration respectively. ( 2 )

Typically the Y-t curve from seismic observation data is like that of below: Figure : The LURR anomaly prior to the Kobe earthquake and the Tottori earthquake.

The LURR reaches to a high value several months prior to the occurrence of the upcoming large earthquake, in the eve of the large earthquake the LURR decrease to a low level and then the large event occurs. The results of LURR in this experiment are shown below. Figure * is the result for G3 (large specimen) and the Figure** is that one for specimen GS2 (small one). Both of them have the common feature that prior to the final fracture of the specimen the LURR reach to a high value, then the LURR decrease and followed by the occurrence of macrofracture. The experimental results coincide with the seismological observation very well. It seems that both the macrofracture and the earthquake have the CP (Critical Point) behavior

G 3 GS 3

AER (Accelerating Energy Release ) Prior to the occurrence of a large or great earthquake the seismic energy release accelerates. In many cases this acceleration can be modeled using a power-law time-to-failure function. The function has a form where E is the cumulative seismic energy, t c is the time of large earthquake, t is the time of the last measurement of E and A, B and m are constants. ( 3 )

In our experiment we focused our attention on the tempo-spacial distribution and evolution of meso-damage. We’ll analyze them in terms of the Statistical Meso-Damage Mechanics later.