A case-based reasoning approach to construction safety risk assessment

Abstract

The fundamental purpose of this research is to encourage continual improvement of construction safety management systems through learning from safety knowledge accumulated in incident cases and past safety plans. To achieve the above purpose, this research developed a novel case-based reasoning approach to risk assessment. The key components of this research are: (1) a generic incident causation model and a specific knowledge representation scheme for abstraction of safety knowledge; (2) an intelligent similarity scoring-based retrieval method that can automatically identify and retrieve relevant knowledge; (3)(a) transparent hazard identification adaptation that can remove irrelevant portions of retrieved cases and integrate the relevant portions into a complete risk assessment tree; (3)(b) risk analysis adaptation that utilises a Poisson model of construction incident occurrence and the Bayesian approach to integrate subjective and object sources to produce more realistic risk estimates; and (4) a prototype system implemented in a realistic case study that validated the proposed approach.