AAU Student Projects - visit Aalborg University's student projects portal
A master's thesis from Aalborg University
Book cover


Elasto-Plastic Constitutive Modelling of Geotechnical Material - Investigation of the Intermediate Principle Stress Influence and Exploration of the General Parametric Yield Surface Format

Translated title

Elasto-plastisk konstitutiv modellering af geotekniske materialer - Undersøgelse af den midterste hovedspændings indflydelse samt udforskning af ``General Parametric Yield Surface Format''

Authors

;

Term

4. term

Publication year

2019

Submitted on

Pages

111

Abstract

Jord og andre friktionsmaterialer beskrives ofte med Mohr–Coulomb- eller Drucker–Prager-flydekriterier. Mohr–Coulomb ser bort fra den mellemliggende hovedspænding, mens Drucker–Prager behandler alle hovedspændinger ens. Virkelig materialeadfærd ligger typisk mellem disse yderpunkter. I praksis bruges Mohr–Coulomb ofte som en konservativ grænse, især i plan deformation, hvor friktionsvinklen mange steder øges empirisk. I afhandlingen undersøges et generelt format for flydeflader, General Parametric Yield Surface Format, foreslået af Lars Damkilde. En flydeflade er den grænse i spændingsrummet, der adskiller elastisk opførsel fra permanent (plastisk) deformation. Tre nye flydeflader, der kan omfatte flere eksisterende modeller, formuleres og implementeres i et numerisk setup. Ved vurdering af plastiske materialemodeller ser vi på nøjagtighed (afvigelse fra data), beregningstid og robusthed. Den første flydeflade har fire parametre; to kan sættes ud af kraft, så modellen falder tilbage på Mohr–Coulomb. Den er beregningseffektiv, men mangler robusthed, især i tredimensionelle analyser. Den anden flydeflade er en glat, kontinuert tilnærmelse til den første. Den opnås med et nyt princip om lokal hjørneafrunding, som efter forfatternes kendskab ikke er beskrevet i litteraturen. Modellen muliggør også en glat afskæring i træk og kan tilpasses optimalt, når der foreligger forsøgsdata. Den viser sig numerisk robust. De to første modeller antager lineær afhængighed af middelspændingen (hydrostatisk spænding). Den tredje model indbygger Boltons udbredt accepterede, ikke-lineære middelspændingsafhængighed for almindelige sandtyper. Den arbejder med Boltons velkendte parametre, giver den mest målrettede materialekalibrering i afhandlingen og viser formatets generelle anvendelighed. Modellerne er implementeret i MATLAB og som Fortran-kode til brug i Abaqus. De er kalibreret mod data fra ægte triaxialforsøg og anvendt i en række elastoplastiske finit element-analyser af typiske geotekniske problemstillinger for at undersøge indflydelsen af den mellemliggende hovedspænding og den numeriske ydeevne. Resultaterne viser et betydeligt uudnyttet potentiale i forhold til Mohr–Coulomb i plan deformation og en mærkbar forbedring i generelle 3D-analyser. En højere bæreevne kan opnås, hvis modellen kalibreres med både triaxiale træk- og trykforsøg—noget Mohr–Coulomb ikke kan håndtere. Den hjørneafrundede model koster mere beregningstid, men er robust, især i 3D, og anbefales derfor, da robusthed er central i numerisk praksis.

Soils and other frictional materials are often described using the Mohr–Coulomb or Drucker–Prager yield criteria. Mohr–Coulomb ignores the intermediate principal stress, while Drucker–Prager treats all principal stresses equally. Real material behavior typically lies between these extremes. In practice, Mohr–Coulomb is widely used as a conservative limit, especially in plane strain, where the friction angle is often increased empirically. This thesis explores a general format for yield surfaces—the General Parametric Yield Surface Format—proposed by Lars Damkilde. A yield surface is the boundary in stress space that separates elastic behavior from permanent (plastic) deformation. Three new yield surfaces, capable of encompassing several existing models, are formulated and implemented in a numerical framework. When evaluating plasticity models, we consider accuracy (discrepancy), computation time, and robustness. The first yield surface has four parameters; by omitting two, it reduces to the Mohr–Coulomb model. It is computationally efficient but lacks robustness, particularly in three-dimensional analyses. The second yield surface is a smooth, continuous approximation of the first, built using a new idea of local corner rounding, which, to the authors’ knowledge, is not otherwise described in the literature. It also enables a smooth tension cut-off and can be optimally fitted when experimental data are available. This model proves numerically robust. The first two models assume linear dependence on the mean (hydrostatic) stress. The third model incorporates Bolton’s widely accepted nonlinear mean-stress dependence for common sands. It operates with Bolton’s well-known parameters, provides the most targeted material calibration in the thesis, and demonstrates the generality of the format. The models are implemented in MATLAB and as Fortran code for use with Abaqus. They are calibrated to data from true triaxial experiments and used in a series of elasto-plastic finite element analyses of typical geotechnical problems to assess the influence of the intermediate principal stress and computational performance. The results show substantial untapped potential compared with Mohr–Coulomb in plane strain and notable improvements in general 3D analyses. Higher bearing capacity can be achieved if the model is calibrated with both triaxial extension and compression tests—something the Mohr–Coulomb model cannot accommodate. The corner-rounded model has a higher computational cost but is robust, especially in 3D, and is therefore recommended because robustness is essential in numerical applications.

[This abstract was generated with the help of AI]