Article Preview
TopIntroduction
The geotechnical and shallow surface seismic analyses are carried out by knowledge of the subsurface profile or stratigraphy of the site under study. An ideal identification of a soil profile for a seismic analysis must be extended to the rock, defined as material with a shear wave velocity greater than 700 m/s, and the physical properties of the soil between the ground surface and bedrock should be defined. Efficient geotechnical recognition and investigation of the subsurface require the exploration of the site under study. This begins generally with a thorough review of the available information about the site, viz. local geology, topographic maps, faults maps, and depth-to-bedrock maps. These data can support inferences about subsurface conditions. Such inferences, however, are rarely sufficient for site-specific design or evaluation, and must be additionally confirmed. Subsurface investigations, accomplished by trenching, drilling and sampling and in situ testing, can provide quantitative information, frequently required in the evaluation of site responses. Logging and sampling should take place at sufficient intervals to detect weak zones or seams that could contribute to ground failure (Kramer & Stewart, 2004). However, these classical techniques of investigation (drilling and sampling), in situ tests, or geophysical means are generally costly and needing heavy equipment and qualified personnel. To avoid these constraints, system identification and inverse problem analyses are used and offer ability to estimate soil properties without the measurement process disturbing the soil mass. Identification theory and inverse problem analyses have been largely documented by a number of published books and papers (Fletcher, 1980; Nelles, 2001; Dahlquist, 1974; Ogunfunni, 2007; Kozin & Nathe, 1986; Pearson, 2004; Zentar et al., 2001). System identification and inverse problem analyses play an important role in development, validation and calibration of soil models, as well as estimation of in situ properties and parameters, using experimental and recorded earthquake data. We are interested here on the role which they play on the characterization and modeling of geotechnical systems. Zeghal and Oskay (2002) developed a system identification technique to identify local soil characteristics and properties of soil-systems using the acceleration records provided by local instrument arrays. They calibrated and evaluated an optimal model of soil response by minimizing discrepancies between recorded and computed accelerations. Tsai and Hashash (2008) presented a review of inverse analysis techniques applied to downhole array data and developed an inverse analysis framework by using downhole array measurements to extract the underlying soil behavior and developed a neural network-based constitutive model of the soil. Oskay and Zeghal (2011) have also summarized previous works on the identification technique used to estimate soil properties from strong motion records. Harichane et al. (2005) and Harichane (2005) proposed a new approach using system identification theory and free field records, for identifying simultaneously soil profile characteristics of two sites. The proposed new approach is based, firstly, on a formulation of a theoretical transfer function or soil amplification function for two sites in terms of the different parameters of the soil profile layers (thickness, damping ratio, shear wave velocity, and unit weight). The soil amplification function is formulated with the assumption of a vertical propagating shear wave through horizontally stratified soil layers of infinite side extent. One dimensional analyses of soil response are extensively used for their simplicity (Govinda Raju et al., 2004). In other hand, transfer functions, or ratios of the Fourier amplitude spectra of input and output acceleration couples, have been widely used to estimate natural frequencies of vibration and associated wave propagation velocities of sites, earth dams, and other systems (Oskay & Zeghal, 2011). In the new approach (Harichane et al., 2005), the amplification function is smoothed with its analogous one obtained from experimental data (spectral ratios) by means of least squares minimization technique according to the Levenberg-Marquart algorithm. The identification of the parameters is performed by solving numerically a non linear optimisation problem. The numerical efficiency and the validity of this procedure have been demonstrated by Harichane et al. (2005) for a single soil profile with experimental data recorded within the Garner Valley Downhole Array (GVDA) (Archuleta et al., 1992) with selected acceleration records at free surface and 22m depth. The major objective of the present study is to provide an experimental validation of the approach by comparing numerical results with those obtained by in situ and laboratory tests. Strong ground motions data recorded during the 2003 May 21, Boumerdes earthquake (Algeria) are used for the purpose.