A probabilistic hybrid sensor fusion and optimization approach for aircraft composite components

An integrated sensor system that continuously monitors the structural integrity of an aircraft’s critical composite components can have a high payoff by reducing risks, costs, inspections, and unscheduled maintenance, while increasing safety. Hybrid sensor networks combine or fuse different types of sensors. Optimal sensor fusion tries to find the optimal number and location of different types of sensors such that their combined probability of detection is maximized. Optimal hybrid sensor networks can be more robust, more accurate, and/or cheaper than networks consisting only of homogeneous sensors. A generic sensor fusion approach that combines the probabilities of detection of heterogeneous sensors is described. A fast greedy optimization approach that provides approximate solutions is described and demonstrated. Computable lower and upper bounds of a probability of detection objective function were determined. Fiber Bragg grating sensors can be inserted in layers of composite structures to provide local damage detection, while surface-mounted piezoelectric lead zirconate titanate sensors can provide global damage detection for the host structure under consideration. The generic approach is demonstrated on such combinations of fiber Bragg grating and lead zirconate titanate sensor networks. It is demonstrated that the proposed approach can be used to answer structural health monitoring network design problems such as the following: (1) Given a number of sensors, what is the maximum probability of detection that the sensors can attain and where should they be positioned to provide the maximum probability of detection? (2) If a given probability of detection is desired, the minimum number, types, and locations of sensors that are needed to attain this probability of detection can be determined. The approach is generic, that is, it can be extended to any number or types of sensors for which probabilities of detection can be defined.