A Bayesian approach to functional sensor placement optimization for system health monitoring

Prognostics and health management of a complex system require multiple sensors to extract required information from the sensed environment and internal conditions of the systems and its elements. A critical decision, particularly in the context of complex systems, is the number and location of the sensors given a set of technical and non-technical constraints. This paper provides a Bayesian Belief Network (BBN)-based sensor placement optimization methodology. The approach uses the functional topology of the system, physical models of sensor information, and Bayesian inference techniques along with the constraints. Utility functions are used for optimized sensor placement based on the value of information that each possible sensor placement scenario provides. We also attempt to answer questions such as: how to infer the health of a system based on limited number of sensor information points at certain subsystems (upward propagation); how to infer the health of a subsystem based on knowledge of the health of main system (downward propagation); and how to infer the health of a subsystem/component based on knowledge of the health of other subsystems/components. Dynamic BBN is used as the engine of projecting the health of the system.