Adobe structures in earthquake zones

Abstract

The aim of this thesis is to apply a reparation method, developed by the Pontificia Universidad Católica del Perú, on an adobe structure damaged by earthquake loading. This loading is applied with a shaking table test. Using these tests an ATENA- GiD model is made which is able to simulate the behavior of adobe houses under earthquake loading. A shaking table test is done to damage an adobe structure, after reparation, the structure is tested again and the test data of the original and repaired building is compared. The ATENA-GiD model is compared with the first shaking table test. The adobe model is made similar to earlier tests performed at the PUCP. The masonry is made with adobe blocks and mortar. The model is dried for 28 days and then subjected to an earthquake loading derived from the May, 1970 earthquake recorded in Lima. The tests consisted of 2 phases, the first phase with a maximum displacement of 30 mm and the second phase with a maximum displacement of 60 mm. Due to this loading, cracks became visible in the structure. These cracks are visible after a second phase, with a maximum displacement of 60 mm. To repair these cracks they are opened and cleaned with a drill and a hammer and pin. After opening, the cracks are repaired using two different methods, the “manual method” and the “silicone method”. The “manual method” consists of manually filling the cracks with the reparation mortar. The “silicone method” is a method where cracks are first covered with silicon and after hardening of the silicon, the reparation mortar is injected in the cracks. After 28 days drying, the structure is tested again with the same signal. Cracks are already visible in the first phase. In the second phase the structure partially collapsed. This is underlined by the data from the two tests. The natural frequency of the walls decreases during the first test. This implies that there is damage in the walls. After repair, the natural frequency goes up, but never reaches the original value. This implies the presence of non-visual damage. In the second test the roof detaches from the walls in an early stage of the test, this due to the non-visible damage that is not repaired and have weakened the structure. The detachment of the roof results in an structure with less stability and strength. Dynamic tests are expensive and time consuming, therefore It is important to find out what finite element software (FEM) can contribute to the research in adobe buildings in earthquake areas. In the recent past, FEM software was used on a very basic scale at the PUCP. With this thesis it is intended to explore the possibilities of using FEM software in the research of adobe buildings. This to further improve the research possibilities on adobe structures with respect to earthquake loading. The model for this thesis is made in ATENA-GiD. Due to the recent development of the scripts for this program, this study is one of the first trails for the program. Because of this, several bugs were detected. During the course of this thesis improvements have been made to the program due to close contact with the developers and the author. In spite of these improvements, the bugs in the program made it only possible to make a very basic model. This model has a coarse mesh and the material is modeled on a macro scale which means that the blocks, mortar and block-mortar interface are smeared out in the continuum. After repair the structure does not have the same behavior as the original test. Due to non visible damage that is not repaired, the structure loses strength and stability after the first test. Besides this, the repaired part is not as strong as the original masonry. Cracks between the longitudinal and transversal wall are very dangerous. The whole construction looses its stability and strength when these cracks are visible. ATENA-GiD can simulate the behavior in the elastic phase. The inelastic phase does not have similar results as with the PUCP tests. Improvements are made to ATENA-GiD throughout the course of this thesis. Due to lack of these improvements this model has a very coarse mesh, which was needed to reduce calculation time. Several updates in the program can now improve calculation time and therefore more time consuming models can be made, which will calculate the inelastic behavior better.