Getting finite element programs to reason about their analysis assumptions

Despite the spectacular improvements of finite element analysis programs, there has been no corresponding progress in the development of physical modeling and interpretation aids. The major bottleneck that hinders such development is the fact that the present, purely numerical programs do not have a higher level understanding of the models they analyze. In particular, they are not cognizant of the physical objects that the models simulate, or of the assumptions built into those models. This paper describes mechanisms for imparting some basic reasoning capabilities to finite element analysis environments, including the ability to generate models by applying suitable assumptions, to evaluate these assumptions and modify them when they ceases to be valid, as well as to track the evolution of the models as more detailed analyses are performed.