The thesis mainly deals with predicting the lateral failure pressure of masonry panels which carry little or no axial loading. This type of isotropic and orthotropic panels can be found as cladding for framed construction or upper floors of multi-storey load bearing structures. The failure criterion has been established recently for an orthotropic brittle masonry panels, but no such criterion exists for isotropic panels. In the first phase of the work, three mortar cross beams, each of 5 different aspect ratios have been tested to establish the failure criterion in biaxial bending for isotropic material. The specimens made of isotropic material cracked and failed simultaneously without shedding of load in the stronger direction whereas in case of orthotropic material, cracking was followed by 'load shedding' to the stronger direction and a subsequent failure of the specimens. The failure criterion that was developed for the isotropic material was incorporated in a finite element plate bending program to predict the failure pressure of panels subjected to lateral loading. In order to test the theoretical model, few isotropic panels with different boundary conditions were tested. This include tests carried out on a total of 4 panels that are simply supported on three sides and four sides under idealised boundary conditions with no rotational restraints. In case of three sides supported panels, there was no difference whether it is free at the top or at one of the vertical sides as the properties were same in both directions and hence only one type of panel was tested. A good agreement between the theoretical and experimental results was observed. The second phase of the project was to develop a quick and reliable method of design by predicting the failure pressure of laterally loaded masonry panels. Artificial neural networks, as a computational tool, offers very exciting technique for computing non-linear engineering problems, similar to the current situation that exists for the design of laterally loaded masonry panels. A hybrid system that combines the capabilities of artificial neural networks and case based reasoning is developed in this thesis. In the present study, multilayered feedforward net with back propagation algorithm is used. To develop this application, a neural network program was developed in C++ incorporating back propagation algorithm and sigmoid activation function. An excellent user interface for this program was developed using Microsoft Foundation Class (MFC) libraries.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:657428 |
| Date | January 1999 |
| Creators | Mathew, Anu |
| Publisher | University of Edinburgh |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Source | http://hdl.handle.net/1842/12606 |
Page generated in 0.0017 seconds