Combining qualitative model‐based diagnosis and observation within fault‐tolerant systems

Article type: Research Article

Authors: Schiller, Frank^; | Schröder, Jochen

Affiliations: Siemens AG, Automation and Drives, Advanced Group Technologies (A&D GT 4), P.O. Box 4848, D‐90327 Nuremberg, Germany E‐mail: [email protected] | Technical University of Hamburg‐Harburg, Control Engineering Department, D‐21071 Hamburg, Germany E‐mail: J.Schroeder@tu‐harburg.de

Note: [] Corresponding author.

Abstract: The increasing importance of fault‐tolerant technical processes leads to new model‐based approaches. In the approach proposed here, the effects of faults are decreased by additional control commands supplied by a fault‐tolerance algorithm. This algorithm combines qualitative model‐based diagnosis and observation. The main idea is to reach the non‐faulty output and non‐faulty state as accurately as possible even in the presence of faults. The signals and faults are described qualitatively by discrete values, and their relations are formalized by an automaton. This automaton is used in diagnosis to identify faults, in observation to identify the present state, and in simulation to identify the correct output and state. A correction mechanism determines an additional control command which would have led to correct behaviour at the current time point and applies it to the process at the following time point. The approach is illustrated by an example of chemical process engineering.

Keywords: Fault‐tolerance, diagnosis, observation, dynamic systems, qualitative modelling, automata

Journal: AI Communications, vol. 12, no. 1‐2, pp. 79-98, 1999