Using Parametric Dynamical Models to Determine Observability Requirements for Condition Monitoring Systems

Abstract

An effective condition monitoring system depends on good observability of the effects of a fault. For many damage mechanisms that affect the reliability of components in systems, parametric models have been developed that describe the rate at which damage accumulates, based either on empirical or theoretical models of the physics of the damage process. These models can inform the reliability analyst about observability of the damage process, and the effect of time-varying system behavior on the observable damage effects. Simulations of system processes with integral damage models can then be used to determine how overall system processes affect damage accumulation in components, particularly in time-varying processes, how damage accumulation compromises performance of the overall system, and the sensitivity of features extracted from signals with respect to fault detection and identification. A simple framework is described for incorporating damage accumulation into lumped-parameter system models. Examples are given for a material handling pipeline process and for a haul truck tire, which highlights some specific challenges of relating overall process state variables to state variables of a damage process, for choosing sensors for a condition monitoring system.