Stochastic Finite Element Analysis of Portuguese Adobe Masonry

Abstract eng:
Earth is a construction material which has been used since ancient times in many parts of the world according to its local availability, low manufacturing cost, and its need for simple construction techniques. Even though earthen constructions have good thermo-acoustic properties, they typically show a very poor performance under earthquake ground motion. Rammed earth and adobe masonry are the main types of earthen construction. Nowadays, it is estimated that approximately 30% of the world population lives in earthen buildings and this percentage increases up to around 50% in developing countries. Such an information highlights the need for a seismic assessment and strengthening of existing earthen structures. The present study is focused on the mechanical behavior of the traditional adobe masonry (AM) of the Aveiro district, Portugal, where approximately 40% of existing buildings are made of adobe and many of them have a socio-cultural value. Extensive surveys have shown a poor state of conservation of AM buildings, the strengthening of which should be based on a comprehensive knowledge of mechanical properties and behavior. To that aim, a nonlinear finite element (FE) modelling approach is used to simulate the experimental behavior of AM in different boundary and loading conditions associated with axial and diagonal compression tests. The latter are amongst the most common experimental tests used for mechanical characterization of masonry assemblages, particularly to define their macroscopic response to uniaxial compression and shear. Based on statistics for mechanical properties of adobe bricks and mud mortar provided by past experimental tests, a macromechanical model of AM was developed within LS-DYNA software and validated against experimental data. The FE models of two types of specimens subjected to axial compression and diagonal compression, separately, were generated. A comparative analysis between numerical and experimental results, both in terms of force–displacement curves and crack patterns, showed that the FE model was able to reproduce the real behavior of AM in different boundary and loading conditions. Afterwards, a single-parameter sensitivity analysis was performed on each AM model to assess whether and how the AM behavior changes under varying material properties. That analysis was the basis for a probabilistic assessment in which a stochastic FE analysis was carried out. Each material property was assumed to be a spatially-distributed random variable in order to reproduce the high level of inhomogeneity provided by material tests on AM constituents, that is adobe bricks and mortar. A small number of model realizations subjected to axial compression was randomly generated through Monte Carlo simulation technique. Two alternative types of stochastic representation were adopted. The former was a simplified stochastic FE modeling (SFEM) in which the spatial variability of material properties was lumped into single brick units, each of them fictitiously extended to the middle of mortar joints. In the second case, an advanced stochastic FE modeling (ASFEM) strategy was used and consisted in a random generation of material properties for all finite elements. It was found that even a limited number of ASFEM simulations allowed the experimental force–displacement response to be captured.

• Export as: BibTeX | MARC | MARCXML | DC | EndNote | NLM | RefWorks
• Add to your basket