Using STAMP to develop leading indicators
Ths paper describes an approach to using STMP and STPA to derive leading indicators of increasing risk. There are always warning signs before a major accident, but these signs may only be noticeable or interpretable as a leading indicator in hindsight. Before an accident, such “weak signals” are often perceived only as noise. The problem then becomes how to distinguish the important signals from the noise. Defining effective leading indicators is a way to accomplish this goal by providing specific clues that can be monitored. A lot of effort has been spent on trying to identify leading indicators, particularly in the petrochemical industry. Much of the past effort has involved finding a set of generally applicable metrics or signals that presage an accident. Examples of such identified leading indicators are quality and backlog of maintenance, inspection, and corrective action; minor incidents such as leaks or spills, equipment failure rates, and so on. Some depend on surveys about employee culture and beliefs, with the underlying assumption that all or most accidents are caused by employee misbehavior, and include as leading indicators such culture aspects as safety awareness, mutual trust, empowerment, and promotion of safety [Am12]. A large number of proposals for leading indicators outside the petrochemical industry focus on occupational safety rather than system safety, and some are simply a listing of potential hazards, such as lack of safety training; whether there is a lock-out, tag-out policy or a stop-work policy; and whether there are medical facilities on site [HSE03]. In fact, the BP Grangemouth Major Incident Investigation Report suggested that industries may have a false sense of safety performance due to their focus on managing personal safety rates rather than process safety1 [HSE03]. As a result of major accidents in the chemical industry, a concerted and long-term effort has been devoted to identifying leading indicators of risk. Most assume that accidents are caused by component failures and that likelihood of failures should be used to reduce the scope of the search for leading indicators despite the fact that likelihood may often be unknown, and the practice may result in overlooking low likelihood events.
Full Text: PDF