An Empirical Evaluation about Using Models to Improve Preliminary Safety Analysis

Abstract

Context: Safety analysis is an activity of fundamental importance in the development of safety-critical systems (SCS) to ensure that hazardous situations are properly found and mitigated. Such analysis is usually performed after a system requirements specification is available. Therefore, it is then worthwhile to investigate specification techniques to detect their strengths and weaknesses with respect to finding and documenting hazards early in the development process. Objective: In this paper, we investigate similarities and differences in the results of a preliminary safety analysis from requirements specified using Textual Use Cases (TUC) and Business Process Modeling Notation (BPMN). Method: We adopted a controlled experiment as research method using computer engineering students as subjects. Results: The subjects of BPMN group found more accidents, hazards as well as more causes of hazards. Moreover, they have a higher preference for the template used for safety analysis documentation. Conclusions: The use of BPMN to represent the interactions among actors in a system probably lead to the discovery of more accidents and hazards, but more experiments are necessary to test this hypothesis since the results are not statistically significant.