Using Ontologies and Machine Learning for Hazard Identification and Safety Analysis

Safety analysis (SA) procedures, such as Hazard and Operability analysis (HazOp) and Failure Mode and Effect Analysis (FMEA) are generally regarded as repetitious, time consuming, costly, and require a lot of human involvement. Previous efforts have targeted automated support for SA at the design stage of system development. However, studies have shown that the cost of correcting a safety error is much higher when done at the later stages than the early stages of system development. Hence, relative to previous approaches, this chapter presents an approach for Hazard Identification (HazId) based on requirements and reuse-oriented safety analysis. The approach offers a convenient starting point for the identification of potential system safety concerns from the RE phase of development. It ensures that knowledge contained in both the requirements document, and previously documented HazOp projects can be leveraged in order to attain a reduction in the cost of SA by using established technologies such as ontology, case based reasoning (CBR) and natural language processing (NLP). The approach is supported by a prototype tool, which was assessed by conducting a preliminary evaluation. The results indicate that the approach enables reuse of experience in conducting safety analysis, provides a sound basis for early identification of system hazards when used with a good domain ontology, and is potentially suitable for application in practice by experts.