Guidelines for the verification and validation of expert system software and conventional software: Survey and documentation of expert system verification and validation methodologies. Volume 3

This report is the third volume in the final report for the Expert System Verification and Validation (V&V) project which was jointly sponsored by the Nuclear Regulatory Commission and the Electric Power Research Institute. The ultimate objective is the formulation of guidelines for the V&V of expert systems for use in nuclear power applications. The purpose of this activity was to survey and document techniques presently in use for expert system V&V. The survey effort included an extensive telephone interviewing program, site visits, and a thorough bibliographic search and compilation. The major finding was that V&V of expert systems is not nearly as established or prevalent as V&V of conventional software systems. When V&V was used for expert systems, it was almost always at the system validation stage after full implementation and integration usually employing the non-systematic dynamic method of {open_quotes}ad hoc testing.{close_quotes} There were few examples of employing V&V in the early phases of development and only weak sporadic mention of the possibilities in the literature. There is, however, a very active research area concerning the development of methods and tools to detect problems with, particularly, rule-based expert systems. Four such static-testing methods were identified which were not discovered in a comprehensive review of conventional V&V methods in an earlier task