Intelligent Agents for Corrosion Prevention in New Vehicle Design

This paper describes the development of an expert tool for vehicle design that assists the designer in avoiding bad design decisions, which may cause corrosion problems during the life of the vehicle. The tool is an autonomous agent installed as part of designer's CAD/CAE system. It uses a Hybrid Expert Production System containing Rules in its Knowledge Base and both an Inference Engine that compares these stored Rules with the information (Facts) provided by the designer and CAD/CAE software and a Bayesian Network for processing uncertainty in the facts and rules. An important enhancement in this system not found in most Expert Systems is an Uncertainty Processor. The Bayesian Net calculates the cumulative uncertainty determined by propagating and combining the estimated error found in each and every Rule and Fact used in the chaining logic required in arriving at the proposed final conclusion. Also included is a Knowledge Acquisition Facility used by corrosion and vehicle design experts to load supporting knowledge and Rules into the Knowledge Base, and an ondemand Explanation Facility to detail the logic used by the system to arrive at the conclusion. The model is verified by simulating the galvanic corrosion of a truck door latch during Accelerated Corrosion Testing.