Original Date: 01/23/1995
Revision Date: 01/18/2007

Performance assessment is predicated on how a system is expected to perform. Probabilistic methods are traditionally used to extract mathematical models for system risk assessments. These models are dependent on a known probability distribution and work best for deterministic (measurable or modelable) situations.

The System Studies Department and the Assessment Technology Department of Sandia are developing methods for using fuzzy logic to address unexpected or abnormal-environment assessment. Some causes of these uncertainties can be natural variability, measurement variability, sampling variability, reporting biases such as human interpretation and selection, and incomplete knowledge.

The utility of fuzzy logic arises from the fact that fuzzy measurement is based on opinion, not on quantitative measurement. Fuzzy uncertainty requires no assumptions about probability distributions, and fuzzy mathematics are simple and direct. These attributes make it applicable to such tasks as safety assessment where warning of undesired and unexpected performance features are critical. Fuzzy estimates require less information about characteristics at the tails (extremes of the variability) which is often not available.

Tools are being developed by Sandia for performing fuzzy/probabilistic hybrid analysis and generation of three-dimensional visual images to graphically demonstrate probabilistic estimates. Sandia will continue investigation of representations for fuzzy and hybrid knowledge, to investigate and develop methods for measuring scale factors, to incorporate variable scale factors, and to construct a complete mathematical assessment.

For more information see the Point of Contact for this survey.