Robust algorithm for optimal configuration of a multi-type network of sensors

An optimal configuration of sensors that maximizes the insight on the structure is essential for reliable condition monitoring. This work presents an approach for optimally placing sensors of different types. The objective of the optimization process doesn't only search for sensor locations that maximize the information gain of each sensor type, but rather builds a network of sensors that collectively reduce the estimation error as their signals are fused to reconstruct the structure's state. Fitness criterion is derived for each sensor type by extending the classical effective independence approach. Based on that, measures of system's observability are used as a metric of information gain and uniqueness for the respective sensor arrangements. Numerical studies are presented to demonstrate the robustness and effectiveness of the method in finding the optimal configuration for an electric motor and a steel structure operating under typical loading conditions and anomalies.