Automated design data capture using relaxation techniques

Abstract

A method that converts a Computer Aided Design (CAD) database description of a process control system into a frame-oriented knowledge base for an automated diagnostic control system is introduced. The system is unique in that relaxation labeling techniques borrowed from research in computer vision are used to resolve the functional composition of the diagnostic frames. The target system for the automated knowledge generation process uses frames to represent structure and function data at the component level and does not use inference rules. Failure detection is performed by constant analysis of the output functions. Diagnosis of failure is performed via functional inversion of the component transfer function represented in the component frame. Capture of the functional knowledge of a component is the most difficult task in the building of this type of knowledge base. A CAD database constructed for a particular process control system contains all the structural information required to describe the system, but little or no functional data. A functional description can be inferred in some cases from the CAD data, however, the complexity of such a system for a domain as large as process control is extensive.

References

1. Jamieson, J. R., et al., (1984). "NASA's LOX Expert System," in Proceedings of American Institute of Aeronautics and Astronautics, Computers in Aerospace V, New Orleans.

Index Terms

• Computing methodologies

  • Artificial intelligence

    • Knowledge representation and reasoning

      • Ontology engineering

  • Machine learning

    • Learning settings