convenient way to represent the behavior of the foundation soils (Chore et al. 2010). Though fourparameter
Burger model represents the soil behavior more realistically in comparison to other lowerorder
models, the use of this model for predictive purpose is not frequent, probably due to the
difficulty in determining the parameters. Dey and Basudhar (2010) demonstrated the efficacy of
four-parameter Burger model in predicting the response of viscoelastic foundation beds, and
validated the same with the aid of a case study. However, the article reported a trial-and-error
scheme to determine the model parameters (two elastic and two viscoelastic parameters). For known
values of model parameters, a forward analysis can easily predict the response of the system. On
the contrary, the determination of the model parameters from the inverse analysis of the observed
system response is complex. Very often, such inverse analysis problems are ill-posed characterized
by the presence of several solutions giving identical responses. Hence, the determined model
parameters become the artifacts providing the response of the system as observed, although, in some
cases they might not have physical significance. An iterative inverse-analysis technique combined
with optimization algorithm may provide satisfactory solution to the above problem.
Conventionally, in a viscoelastic model, coefficients of viscous dashpots are functions of timedependent
parameters of soil, such as the degree of consolidation. Several researchers have made
use of arbitrary dashpot coefficients in the formulation and analysis (Freudenthal and Lorsch 1957,
Hoskin and Lee 1959, Murayama and Shibata 1961, Freudenthal and Spillers 1962, Schiffman et al.
1964, Bandopadhyay 1982); however, they did not suggest any generalized technique for the
reasonable determination of the model parameters. Few simple techniques have been adopted by
different researchers for simpler models. Taylor (1942) proposed Theory-B of consolidation and
employed the ratio of incremental pressure to the induced settlement, to determine the coefficient of
viscosity of a homogeneous viscoelastic medium using the data obtained from laboratory
consolidation test. Lo (1961) performed long-term consolidation tests on remolded and undisturbed
clay samples to determine the time-settlement behavior of cohesive soils due to secondary
consolidation and proposed two different methodologies to determine the rheological parameters for
a three-parameter model. The first method uses a ratio of incremental settlement with respect to
incremental stress. The second method uses a plot of logarithmic incremental settlement to time,
subjected to a constant incremental time and a constant viscoelasticity ratio. Schultze and Krause
(1964) carried out large-scale compression tests to determine the magnitude of the residual pore
water pressure, and developed isochronous curves for various time durations to provide qualitative
rheological prediction about the viscosity of clayey soil. Juskiewicz-Bednarczyk and Werno (1981)
presented the determination of one-dimensional consolidation parameters for a viscoelastic medium
represented by a three-element lumped parameter mechanical model. The coefficient of viscosity
and other consolidation parameters were determined with the aid of an iterative procedure applying
minimization of an error function with respect to the unknown parameters. It is recognizable that
the applicability of the above techniques becomes difficult with the increase in the complexity of
the model. At the same time, complexity of a model, to a limited extent, is required, for the proper
representation of the constitutive behavior of the medium; for e.g. at least four-element lumped
parameter model is required for the proper representation of a viscoelastic medium (Dey and
Basudhar 2010). Under such conditions, adoption of inverse analysis, aided with modern computing
techniques applying optimization and neural networks, to determine the parameters from the stressdeformation-time
response of a viscoelastic medium from laboratory or in-situ investigations and
likely to be more computationally efficient.
This paper reports the development of a generalized methodology to determine the model