Optimization framework for calibration of constitutive models enhanced by neural networks

Abstract

SUMMARYA two-level procedure designed for the estimation of constitutive model parameters is presented in this paper. The neural network (NN) approach at the first level is applied to achieve the first approximation of parameters. This technique is used to avoid potential pitfalls related to the conventional gradient-based optimization techniques, considered here as a corrector that improves predicted parameters. The feedforward NN (FFNN) and the modified Gauss-Newton algorithms are briefly presented. The proposed framework is verified for the elasto-plastic modified Cam Clay model that can be calibrated based on standard triaxial laboratory tests, i.e. the isotropic consolidation test and the drained compression test. Two different formulations of the input data to the NN, enhanced by a dimensional reduction of experimental data using principal component analysis, are presented. The determination of model characteristics is demonstrated, first on numerical pseudo-experiments and then on the experimental data. The efficiency of the proposed approach by means of accuracy and computational effort is also discussed.

Selection of sand models and identification of parameters using an enhanced genetic algorithm

“…); (iii) neural networks algorithms (Ghaboussi and Sidarta ; Obrzud et al . ); and (iv) particle swarm optimizer algorithms (Knabe et al . ).…”

A combined neural network/gradient‐based approach for the identification of constitutive model parameters using self‐boring pressuremeter tests

“…Such approach has already been verified and presented in detail in [15]. Hereafter, the strategy is adopted for the SBPT problem in order to identify the MCC model parameters.…”

“…Consequently, the optimized values of R p may tend to be higher than those evaluated through laboratory tests. However, the comparative analysis may be less reliable because the estimates of R p for laboratory tests were obtained through the correlation presented in Equation (15). The numerically optimized values of R p may also sustain the geometry effect since all of the analyses were carried out with the simplified numerical mesh representing a unit of pressuremeter membrane.…”

“…§ The NNs, as far as in situ applications are concerned, were used to derive the overconsolidation ratio [35] or the undrained shear strength from piezocone penetration data [36], to predict movements of natural slopes [37] or to evaluate deflections of diaphragm walls [38], and finally to map geophysical measurements of the complex permittivity of a soil-water electrolyte system onto soil properties such as water content and degree of saturation or density [39]. In the context of parameter identification, the § A brief review of soft computing methods that involve NNs in geomechanics is presented in [15]. 826 R. F. OBRZUD, L. VULLIET AND A. TRUTY NN approach was also explored for laboratory tests such as identification of soil compaction characteristics [40] or identification of elastic properties for orthotropic materials [41].…”

“…In this paper, we present a further development of the approach proposed in [15] with special reference to the complex BVP of SBPT. Hereafter, neural networks (NNs) are used to map experimental data of pressuremeter tests onto close approximates of characteristics of the modified cam clay (MCC) model.…”