Managing qualitative simulation in knowledge-based chemical diagnosis

Authors

  • James K. McDowell,

    1. Dept. of Chemical Engineering and Laboratory for AI Research, Ohio State University, Columbus, OH 43210
    Search for more papers by this author
  • James F. Davis

    Corresponding author
    1. Dept. of Chemical Engineering and Laboratory for AI Research, Ohio State University, Columbus, OH 43210
    • Dept. of Chemical Engineering and Laboratory for AI Research, Ohio State University, Columbus, OH 43210
    Search for more papers by this author

Abstract

Deep knowledge about process behaviors plays an important role in the diagnosis of chemical processes. Cause-and-effect reasoning using deep knowledge is useful especially for interacting malfunctions. This work explores the integration of deep knowledge into task-specific, knowledge-based architectures for resolving interacting multiple malfunctions and presents a novel methodology called diagnostically focused simulation (DFS). Invoked in an auxiliary manner, DFS uses deep knowledge and performs qualitative simulation in a highly constrained manner. The close integration with other problem solvers is an evolutionary approach to using qualitative simulation in diagnosis and manages a normally computationally-explosive procedure. Diagnostic results from the compiled problem solver provide a situation-specific assessment of the chemical process, identify possible malfunction scenarios, and focus on appropriate levels of process detail. DFS effectively demonstrates a balance between run-time simulation and compiled problem solving in diagnosis.

Ancillary