Trust-region reflective optimisation to obtain soil visco-plastic properties

Publication Type:
Journal Article
Engineering Computations (Swansea, Wales), 2016, 33 (2), pp. 410 - 442
Issue Date:
Filename Description Size
EC-11-2014-0236.pdfPublished Version1.64 MB
Adobe PDF
Full metadata record
© Emerald Group Publishing Limited 0264-4401. Purpose - A non-linear creep function embedded in an elastic visco-plastic (EVP) model can simulate the decrease of creep compression rate with time. It overcomes the limitation of a linear logarithmic creep function, by which creep continuously increases to infinite value as time approaches infinity. However, the determination of the creep model parameters is a challenging task to apply the EVP model. Therefore, the purpose of this paper is to present a new numerical solution to determine the EVP model parameters applying trust-region reflective least square optimisation algorithm and the finite difference scheme to simulate stress-strain behaviour of soft soil. Design/methodology/approach - In this paper, the developed method is verified against the field case study of Vsby test fill. A set of EVP model parameters is obtained by applying the developed method to the available laboratory consolidation results of Vsby clay. Then, the predictions of settlement and the excess pore water pressure at different depths are compared to the available field measurement. Findings - The analysis results show the developed method is a reliable tool to evaluate the long-term performance of soft soils under embankments. Practical implications - Practicing engineers can use the proposed optimisation algorithm to increase the accuracy of the soil visco-plastic model parameters by utilising all laboratory results of several loading stages during and after the dissipation of the excess pore water pressure. Originality/value - A novel numerical solution to determine the visco-plastic model parameters applying trust-region reflective least square optimisation algorithm has been developed.
Please use this identifier to cite or link to this item: