On-line fault monitoring of a class of hybrid systems using templates with dynamic time scaling

Abstract

Fault monitoring is the online analysis of process observations to determine if they correspond to correct process operation. In automated manufacturing systems, the observed input and output signals can commonly be characterized as observed time functions of discrete events, and fault monitoring can only rely on the analysis of the timing and sequencing of these events. In many situations, these timing and sequencing relationships are not fixed, but rather depend upon underlying continuous state dynamics of the system.

In this paper, we examine the issue of on-line fault monitoring of a class of hybrid dynamical systems in which the evolution of a continuous state is observed through and influenced by discrete events. Specifically, we consider a process for which the correct behavior is defined by the concurrent operation of any number of instances of a given hybrid automaton, where the hybrid automaton has a special structure formed from the composition of the discrete state structure of the actuators with the continuous state variable. We develop a distributed method for examining the discrete event observations from the process and determining on-line whether they are consistent with the model of correct behavior. The method presented, dynamically scaled templates, is an extension of the template monitoring method developed for monitoring of automated manufacturing systems.

L. E. Holloway has been supported in part by Rockwell International, NSF grant ECS-9308737, NASA grant NGT-40049, and the Center for Robotics and Manufacturing Systems at the University of Kentucky.