Back analysis of model parameters in geotechnical engineering by means of soft computing

Abstract

SUMMARYIn this paper, a parameter identiÿcation (PI) method for determination of unknown model parameters in geotechnical engineering is proposed. It is based on measurement data provided by the construction site. Model parameters for ÿnite element (FE) analyses are identiÿed such that the results of these calculations agree with the available measurement data as well as possible. For determination of the unknown model parameters, use of an artiÿcial neural network (ANN) is proposed. The network is trained to approximate the results of FE simulations. A genetic algorithm (GA) uses the trained ANN to provide an estimate of optimal model parameters which, ÿnally, has to be assessed by an additional FE analysis. The presented mode of PI renders back analysis of model parameters feasible even for largescale models as used in geotechnical engineering. The advantages of theoretical developments concerning both the structure and the training of the ANN are illustrated by the identiÿcation of material properties from experimental data. Finally, the performance of the proposed PI method is demonstrated by two problems taken from geotechnical engineering. The impact of back analysis on the actual construction process is outlined.

Optimization framework for calibration of constitutive models enhanced by neural networks

“…Plastic strains occur when the stress state reaches the state boundary surface given by Equation (15) and the flow rule is assumed to be associateḋ…”

“…Typically, statistical methods can be used to solve an inverse problem. The mixed formulation of genetic algorithms and NNs [15] or the stochastic approach [25] can be found in the literature. Recently, genetic algorithms have also been adapted to solve the model calibration problems, cf.…”

A hybrid finite element and surrogate modelling approach for simulation and monitoring supported TBM steering

“…If optimization algorithms such as Particle Swarm Optimization (PSO) [13] are used for inverse analyses, often a large number of realizations is required. Since this is connected with a prohibitively large effort if large-scale 3D finite element models are used, often surrogate models (alternatively also denoted as meta models) are employed for the evaluation of the objective functions [14,15]. In [16], this approach is used for back analysis of material parameters and steering of the mechanized tunnelling process.…”

“…In geotechnical problems, ANNs have been applied as surrogate models trained by means of numerical simulations and used e.g. for the prediction of the deformations induced by geotechnical interventions [15] or for the prediction of tunnelling-induced settlements [18,19,20,21,22]. Hybrid surrogate modelling approaches in mechanized tunnelling combining POD and ANNs are presented in [23] and [24].…”

Rheological parameter estimation for the prediction of long-term deformations in conventional tunnelling

“…It has been applied successfully to generalize the basic properties of rock mass from laboratory experiment data (Meulenkamp and Grima, 1999;Sonmez et al, 2006), to determine the model parameters from numerical simulations and monitoring data (Feng et al, 2000;Pichler et al, 2003), to help the design and maintenance of tunnels, embankments, slopes etc. (Cai et al, 1998;Mayoraz and Vulliet, 2002;Pichler et al, 2003;Rangel et al, 2005). The attractiveness of artificial neural network comes from its remarkable information processing capability pertinent to nonlinearity, high parallelism, fault and noise tolerance, self-learning and generalization.…”