Optimal sensor architecture selection for health management of complex systems

With technological progress, humans tend to create engineering systems with constantly increasing complexity and higher operational requirements. Many complex systems require the use of a Health Management (HM) solution to ensure safety and enable lifecycle properties management of the system. HM solutions such as Integrated Vehicle Health Management (IVHM) hinge mainly on data obtained from sensors. Sensors or collections of sensors, forming sensor architectures constitute an important fraction of the cost of an HM solution and thus have to be carefully designed. However, the trade-off between cost and performance of a sensor architecture is not yet well understood. In the light of this, we have developed a Pareto optimal sensor architecture selection tool for dynamic systems, which integrates performance and cost and aims at aiding design decisions. The tool uses a performance metric based on Mean Square Error (MSE) and derived from the observability matrix of an estimated state space model for a nominal system operation as well as for different system failure modes. The tool is applied to a case study involving a ducted fan, which is a dynamic system and a common mechanical set-up used for propulsion applications. This system can exhibit different mechanical as well as electrical failure modes throughout its lifecycle, which can be managed using a sensor architecture. We consider 63 possible sensor architectures (all the possible combinations out of six sensors) and the tool reduces the choice to only 13 Pareto optimal ones.