Eliminating “Unknown Known” From Process Safety Management; Better Protection Against Potentially Large Losses

Abstract for 5^th CCPS Latin American Conference on Process Safety

Suggested session: Practical methods for PSM Improvement

Major accidents such as Texas City, Buncefield and Fukoshima involved events which were either not anticipated or were considered to be so unlikely as to require neither thorough study nor adequate safeguards.  Our concern is that Hazard Identification (using techniques such as Checklists, HAZID or HAZOP) sometimes fails to identify significant scenarios because participants are unaware of past events or consider their recurrence too unlikely in the plant understudy. Such attitudes give rise to a class of hazard called “unknown knowns”; hazards which are unknown to those concerned with a plant but which are known to have occurred elsewhere in the industry. Such ignorance risks repeating others mistakes in the past.

Our paper tries to address this concern by providing facilitators of PSM studies with a tool to quickly reference major accidents according to their initiating events, the equipment involved and the failures in layers of protection which occurred. The tool takes the form of an accident database categorised according to Reason's Swiss cheese model to establish a strong connection between the outcome and the procedures or engineering protections which failed allowing the initial event to escalate into a major incident. 

The paper provides examples of the use of the database during preparation for PSM studies to develop a list of major accidents which have occurred on other plants of the type about to be studied. This information prepares the facilitator to ensure systematic analysis of the protections which prevent the accident recurring on the plant under study.

Another application is outlined in which the database is used to refine frequency assumptions made during Risk Ranking or QRA studies to avoid underestimating the frequency associated with an accident scenario and so discarding it as “not credible” or  “low risk” when accident history shows it has occurred elsewhere.

These case studies illustrate practical ways to learn from past accidents and ensure that lessons learned are considered not only at the Hazard Identification stage but throughout the Process Safety Management programme.