National Technical University of Athens (NTUA)
Constitutive modeling of the monotonic and cyclic response of granular soils with applications in seismic liquefaction
Abstract
dc:descriptionThe accuracy of numerical analyses of boundary value problems (BVPs) in GeotechnicalEngineering depends significantly on the reliability and the proper calibration of theemployed constitutive model, depending on the soil type and the loading conditions. Forgranular soils (like sands and gravels), different constitutive models and/or differentcalibrations of the same model are usually adopted, depending on the loading conditions(e.g., monotonic vs. cyclic, under drained vs. undrained conditions, short and intenseseismic excitation vs. sustained cyclic loading of many cycles of small to mediumintensity). Historically, given its complexity, the simulation of cyclic response of granularsoils has attracted great interest in the literature, with a multitude of publications presentingrelevant constitutive models. Very often, however, publications that present relevantmodels exhibit accurate numerical simulation of the monotonic response and include onlya few, hand-picked, cyclic tests with emphasis on liquefaction. Such a validation procedurecan hardly be considered sufficient for cyclic loading, in general. On the contrary, modelvalidation should cover the whole range of cyclic response in each discrete strain regime.Namely, the validation process should focus distinctly on: a) small strains, where the soilresponse is “quasi-elastic” and the use of the true elastic parameters (e.g., measured fromfield tests) is crucial for an accurate simulation, b) medium strains, where the response isnon-linear hysteretic and accumulation of strains with the number of applied cycles isobserved and, finally, on c) large strains, where of importance are the deformations due toliquefaction and the post-liquefaction strain accumulation. Hence, despite the plethora ofvery remarkable specialized constitutive models in published literature, there are very fewpublications that present model validations against the whole foregoing strain range.Concurrently, it has been observed that some of the models that have exhibited asatisfactory performance for cyclic loading may not be as accurate when it comes tomonotonic loading, or need to be recalibrated for that type of loading. Based on the above,there is a need for a constitutive model for granular soils, which will be able to captureaccurately both the monotonic response (until the critical state) and the cyclic response (forany shear strain level) with a single set of soil - specific parameters for any relative densityand stress level. This is the main purpose of the constitutive model developed in the contextof this Thesis.The new elasto-plastic constitutive model is based on Critical State Theory, itbelongs to the family of SANISAND bounding surface models and it inherits concepts fromthe NTUA-SAND model, albeit modified. Its basic characteristic is the use of the last loadreversal point to define both the elastic and the plastic strain rates. Special attention is givento the proper definition of the last load reversal point, so as to mitigate stress-strainovershooting. The model incorporates a macroscopic cumulative fabric function scaling theplastic modulus and targeting an accurate rate of strain accumulation with loading cycles.Post-liquefaction strains are also in focus with an appropriate modification of the dilatancyfunction. The constitutive model has a total of 14 parameters requiring calibration, 12 ofwhich target the monotonic response and should be calibrated first, i.e., before theremaining 2 that refer solely to cyclic loading. Of these 14 parameters, 9 are calibrated firstand directly on the basis of experimental data, while the remaining 5 parameters (3 formonotonic and 2 for cyclic loading) require trial-and-error runs.The constitutive equations are developed at element level and the calibration ofmodel parameters is presented for five (5) sands from the literature (Nevada, Toyoura,Ottawa - F65, M31 and Monterey) and two (2) gravels. As a whole, the selected soils forcalibration purposes cover a range of mean diameter of grains (D50) from 0.1 to 9 mm. Thevalidation includes comparisons to element test data for different types of loading anddrainage conditions, corresponding to a very large range of initial conditions (whole rangeof relative density Dr = 2 – 90% and initial mean effective stress po = 33.33 - 2000 kPa).When experimental data for a specific soil are lacking, empirical relationships from theliterature are employed. This thorough investigation focuses on all levels of induced cyclicdeformations on the basis of the rationale discussed above, but also satisfies therequirement for a unique set of soil-specific model parameters, regardless of the loadingconditions. The validation process proves that the proposed constitutive model is, bydesign, a general-purpose model that provides satisfactory accuracy without a need forrecalibration regardless of whether the problem is static, cyclic or dynamic, at least whenit comes to sands. When it comes to gravels, the de-activation of fabric-related functions(via simply nullifying the respective Νο model parameter) in monotonic simulations provessufficient, without changing any other model parameter.For its possible use in the numerical analyses of BVPs, the new constitutive model isimplemented in finite difference numerical codes FLAC and FLAC3D. These codes can beused in fully-coupled nonlinear numerical analyses of both static and dynamic boundaryvalue problems of geostructures. The constitutive equations are initially written in C++programming language, and then are implemented into the numerical codes as a UserDefined Model (UDM) routine. For the implementation, a second order Euler integrationscheme with automatic error control and a sub-stepping technique are adopted.At the final stage of the Thesis, the ability of the calibrated constitutive model tosimulate sand response at system level is evaluated. For this purpose and due to theircomplexity, dynamic BVPs including liquefaction are selected. The validation is performedagainst experimental results from nine (9) dynamic centrifuge tests on Nevada sand,grouped in four (4) different categories:(a) the seismic response of a horizontal liquefiable sand layer,(b) the lateral spreading of a mildly sloping liquefiable sand layer,(c) the seismic response of shallow foundation systems in liquefiable sand profiles withdifferent stratigraphies subjected to different seismic excitations,(d) the seismic response of a pile group subjected to the lateral spreading of a mildlysloping liquefiable sand layer.Categories (a), (b) and (c) are analyzed in the two-dimensional (2D) space usingnumerical code FLAC, while category (d) is analyzed in the three-dimensional (3D) spaceusing numerical code FLAC3D. Since the employed liquefiable sand is the same in all theabove centrifuge tests (Nevada sand in different relative densities, Dr = 40 – 85%), thefollowing additional modeling approaches are adopted: (i) a common calibration of theconstitutive model, based on laboratory test results at element level and (ii) a commonanalysis methodology (in terms of adopted permeability coefficient, numerical damping,etc.). In this way, the comparison of simulations to data reflects purely the satisfactoryability of the model to simulate the response of a specific sand material, regardless of theproblem at hand and without any need for recalibration.Finally, some of the aforementioned BVPs are studied parametrically in order toevaluate the sensitivity of the numerical results to the use of two (2) of the novel constitutiveingredients. The focus is on the adopted methodology for the mitigation of stress-strainovershooting and on the incorporated factor affecting the dilatancy coefficient targetingpost-liquefaction shear strain accumulation. The results of the sensitivity analysesunderline that these constitutive features are useful, but possibly not crucial for accuracy,at least for the examined BVPs.
Degree
thesis:*- Grantor dc:publisher
- National Technical University of Athens (NTUA)
- Year dc:date
- 2022
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Limnaiou, Taxiarchoula
- Λημναίου, Ταξιαρχούλα
Subjects
dc:subject × 23- Εδαφομηχανική
- Καταστατική προσομοίωση
- Κοκκώδη εδάφη
- Γεωτεχνική μηχανική
- Γεωτεχνική σεισμική μηχανική
- Μονοτονική φόρτιση
- Ανακυκλική φόρτιση
- Ρευστοποίηση
- Προσομοιώματα οριακής επιφάνειας
- Soil mechanics
- Constitutive modeling
- Granular soils
- Geotechnical engineering
- Geotechnical earthquake engineering
- Monotonic loading
- Cyclic loading
- Liquefaction
- Bounding surface plasticity
- Επιστήμη Πολιτικού Μηχανικού
- Επιστήμες Μηχανικού και Τεχνολογία
- Πολιτική μηχανική
- Civil Engineering
- Engineering and Technology
Rights
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/52587