A rule-based expert system for automatic error interpretation within ultrasonic flaw model simulators

This paper describes the development of a novel rule-based expert system application that automatically runs a set of theoretical models used to simulate test procedures for ultrasonic testing methods in nondestructive evaluation and interprets their results. Theoretical modeling is an essential tool in verifying that the test procedures are fit for their intended purpose of defect detection. Four validated models are available to simulate theoretical ultrasonic flaw modeling scenarios. Under certain conditions, the models may break down and produce warning flags indicating that results may not be considered accurate. A considerable level of expertise in the theoretical background of the models is required to interpret these flags. The expert system addresses any warning flags encountered by adjusting the original simulation parameters and rerunning the test in order to produce a valid simulation. Warning flags are addressed by a rule file, which contains formal rules developed from knowledge-elicitation sessions with suitably qualified engineers. The rule file represents the action an engineer would adopt to counter highlighted warning flags. A description of the system and rule-base design is given as well as how the system and its performance were validated.