Validation of Knowledge‐Based System with Multiple Bridge Rail Experts

Abstract

A procedure is described that can be used to validate a knowledge‐based expert system. The method applies four qualitative measures that are specially designed for problems solved by multiple levels of experts whose authority and expertise can be modeled hierarchically. Development of test cases and performance measures are discussed. The proposed method is applied to a system that selects and ranks bridge rails from a list of candidate configurations. This rule‐based system aids engineers in selecting bridge rails according to conditions of the roadway, bridge, and surrounding environment. Based on the level of agreement with human experts and the Turing test, it is shown that the prototype system exhibits a relatively high level of expertise. Other tests and observations indicate that some system deficiencies exist and several features need modification and refinement.