Dynamic Across-Time Measurement Interpretation: Maintaining Qualitative Understandings of Physical System Behavior

Incrementally maintaining a qualitative understanding of physical system behavior based on observations is crucial to real-time process monitoring, control, and diagnosis. This paper de-scribes the DATMI theory for dynamically maintaining a pinterp-space, a concise representation of local and global interpretations consistent with the observations over time. Each interpreta-tion signifies alternative paths of states in a qualitative envisionment. Representing a space of interpretations, instead of just a "current best " one, avoids the need for extensive backtracking to handle incomplete or faulty data. Domain-specific knowledge about state and transition probabilities can be used to maintain the best working interpretation as well. Domain-specific knowledge about durations of states and paths of states can also be used to further constrain the interpretation space. When all these constraints lead to inconsistencies, faulty-data hy-potheses are generated and then tested by adjusting the pinterp-space. The time and space complexity of maintaining the pinterp-space is polynomial in the number of measurements and envisionment states. ACKNOWLEDGMENTS Foremost, I would like to thank my advisor, Ken Forbus, for his much-appreciated encour