| Abstract
|
Understanding the behavior of a structure … Understanding the behavior of a structure and ist components in thepreliminary design phase can have a significant effect on the quality of thefinal design. Accurate prediction of the behavior of a structure can reducethe number of alternative solutions and avoid costly design revisions.However, there are few tools available for modeling and qualitativelyanalyzing a structure.The report describes a system, QStruc, for qualitative structural analysiswhich combines first principles in structural engineering and experientialknowledge of structural behavior. The purposes of QStruc are (1) to generatequalitative response of the structure in terms of deflected shape, moments,and reactions. The framework consists of two main modules: the modelgeneration module and the qualitative analysis module. The proposedqualitative analysis strategy is a "greedy," depth-first approach that triesto expand the derived response as much as possible from known parametervalues. The system makes use of: Causal ordering mechanism, that enables the system to identify thesolution path for the qualitative analysis. A causal model is used todescribe the behavior of a structure in terms of physical quantities andcausal interactions among them. Qualitative calculus, which enables the qualitative evaluation of thephysical quantities of the causal model that describes the behavior of thestructure. Quantity Lattice, which enables the system to reason about partialordering among physical quantities and to reduce some of the ambiguousconclusions caused by the impreciseness of the information.The report discusses the following aspects of QStruc: (1) representation ofthe physical model, fundamental principles, and experiential knowledge, (2)model generation, and (3) qualitative analysis. We provide a simple exampleto illustrate the operations of the prototype system QStruc. The reportconcludes by summarizing our preliminary findings and suggesting directionsfor future research. suggesting directionsfor future research.
|
