A systems approach to earthquake vulnerability assessment

Abstract

The ability to take decisions about the expected response of existing projects (i.e. buildings, lifelines, cities) to an earthquake is difficult and complex. The behaviour of a few selected parameters of the main structural system (e.g. inter-storey drift) are commonly used to make judgements. The behaviour of a project clearly depends upon the structure but it also depends upon many other factors which often are not considered. These include, safety culture, management, condition, use, construction, materials and so forth. The modelling and measurement of these factors vary in quality since they are very different in nature. A model which enables these factors to be put together to assess the proneness to failure of a particular project is proposed. The model follows a systems approach and concentrates on the modelling and management of information. The management of the uncertainty, which is classified into fuzzincss, incompleteness and randomness, is an important part of the model. Hierarchically arranged holons describe the processes making up the project and capture inherent fuzziness of the problem. The model includes tests (such as audits) which a project must pass in order to be declared dependably safe. Dependability is a measure of the degree to which an engineering theory has been tested in practical problems. The proposed methodology combines existing numerical models as well as ways of processing vague information and expert judgement. It is also a very flexible tool which allows the handling of various types of projects and situations which are slightly different from past experience. Experts will use linguistic assessments to measure the evidence about the dependability of processes to sustain their function during an earthquake. Linguistic assessments are matched to interval probability numbers. An interval number is used to capture, in a practical manner, features of fuzziness and incompleteness. Interval probability theory is used to combine evidential support values throughout the hierarchy. A computer implementation of the model (i.e. EVAS) was developed to show its potential for practical use. The software developed was used to apply the methodology to the Hospital Regional de Buenaventura in Colombia. Further testing of the proposed model and EVAS in practical applications should be carried out to ensure their dependability.