Knowledge acquisition and synthesis in a multiple source multiple domain process context

Abstract This paper is concerned with the application of Expert Systems to the arc welding domain. In particular, for two complementary welding processes, namely, the Self-Shielded Flux-Cored and the Submerged-Arc processes. The paper reviews the phases of a 6-year project and concludes that qualitative reasoning holds out the most potential for combining knowledge from related domains. It is also suggested that conventional approaches to so-called knowledge advancement in a scientific sense may ultimately result in domain-specific information and knowledge that is its own barrier to more general usefulness. With hindsight it would seem that the lessons to be drawn from this project are more to do with misconceptions and misunderstandings about expert systems and the nature of knowledge itself, and about the danger of underestimating the complexity of a domain, more so when it has already been a research topic for a considerable period of time. The discussion is purposely software independent because it is the authors' view that whilst the software and knowledge representation formalisms have a bearing on the ease of solution of the problem, they are comparatively secondary issues. The key driving force for this project has been the advancement of knowledge about the welding domain itself. It would appear that expert systems techniques are not only a suitable approach to the solution of this problem, but that they are also a catalyst for the rigorous and logical scrutiny of the domain knowledge, its gaps, and its inconsistencies. Therefore there is as much to be gained from the process of building an expert system as there is from the final product.