TR040: A Construction Risk Management System: An Expert System for Structural Failure/Accident and Seismic Risk of Buildings

An expert system for evaluating the risk to a construction project at all stages of the project is proposed as a means of mitigating the risks in the Japanese construction industry which has seen rapid growth in recent times. The purpose of such an expert system would be to identify the risks which are likely at different stages of the project from conception, design, construction to operation. Such a system would aid management to identify and evaluate potential risks at all stages of the project and if necessary to take early action so as to have the most impact.

In this research we have concentrated our efforts on some of the aspects of the construction risk management system. We have built prototypes of modules which can assess the risk at the construction stage, the operational stage and seismic risk over the lifetime of the building. The modules have been written in NEXPERT OBJECT, an expert system shell, and the prototype system presently consists of knowledge bases, structural failure and construction accident databases, external subroutines written in Fortran and some user-supplied information. The inputs to the prototype system are information that is interactively entered by the user and the rest of the data is read directly from project databases, which are now becoming increasingly common with the bigger construction companies in Japan.

At the heart of the modules which assess the risk at the construction and operation stages is a fuzzy pattern-matching technique for matching the causes of the project to those found in the instances which experienced structural failure and, or construction accidents. Pattern-matching is used to identify those cases of failure/accident in the failure/accident database which are highly correlated with the project. The seismic risk module consists of an external Fortran program for calculating the stochastic response spectra and procedures based on ATC-13 and ATC-2I for determining the expected damage factor using the concept of structural scores.

The prototype system that we have built, though it is a long way from being implemented for practical use, and needs further refinements, has shown that the technology of expert systems as embodied in an object-oriented frame is a very powerful tool in this field.