부산대학교 도서관

This paper addresses the problem of requirements engineering for complex socio-technical systems where an optimal set of technology components and human operators have to be selected to achieve system goals. Goals are achieved by tasks that are expressed as operational scenarios with variations in environmental conditions. The approach taken is to develop a probabilistic model of system reliability as a Bayesian Belief Network (BBN). The BBN model predicts human and machine reliabilities, given input variables representing the scenario and ranges of environmental conditions (i.e. weather, climate, etc.). A software tool, the System Reliability Analyser (SRA) is described that runs a set of scenarios against the BBN models while systematically varying the ranges of 12 input variables specifying properties of human operators such as training, technical equipment specification and environmental conditions. The tool reports human and technical equipment specifications that "survive" the scenario testing at a reliability level higher than a user-defined level. A case study evaluation of the tool is reported using a naval command and control domain, in which different combinations of human roles and equipment requirements specifications are automatically validated against a set of scenarios describing missile attacks on a navy frigate. The implication of using BBN technology and the SRA tool for automating socio-technical system requirements validation and optimising requirements selection in component-based systems is discussed.