In simulations involving uncertainty, two types of unknowns must be taken into account: (1) known unknowns and (2) unknown unknowns. For known unknowns, the nature of the task is known, adequate historical data is available, and although the value of the model variable is unknown, either a theoretical or an empirical probability density function can be established to describe the variable. For unknown unknowns, the value of the variable can be zero, if the task or event does not actually occur, or may go to any amount (either negative or positive) if the event does occur. For example, there may be no definable upper limit if an in-house activity fails catastrophically or a subcontractor fails to deliver the work. These unknowns invariably result in disruptions to operations and significant cost overruns. In industry today, we are particularly concerned with designing proactive control systems. These "unknown unknowns" therefore cannot be ignored. This paper discusses a methodology to incorporate this second type of unknown into a simulation model. Examples include modeling a forklift-pedestrian collision, a labor strike at a critical supplier, and a natural disaster at a factory.
展开▼
