A sensor fusion strategy based on a distributed optical sensing of airframe deformation applied to actuator load estimation

Real-time health monitoring of mechatronic onboard systems often involves model-based approaches comparing measured (physical) signals with numerical models or statistical data. This approach often requires the accurate measurement of specific physical quantities characterizing the state of the real system, the command inputs, and the various boundary conditions that can act as sources of disturbance. In this regard, the authors study sensor fusion techniques capable of integrating the information provided by a network of optical sensors based on Bragg gratings to reconstruct the signals acquired by one or more virtual sensors (separately or simultaneously). With an appropriate Fiber Bragg Gratings (FBGs) network, it is possible to measure directly (locally) several physical quantities (e.g. temperature, vibration, deformation, humidity, etc.), and, at the same time, use these data to estimate other effects that significantly influence the system behavior but which, for various reasons, are not directly measurable. In this case, such signals could be “virtually measured” by suitably designed and trained artificial neural networks (ANNs). The authors propose a specific sensing technology based on FBGs, combining suitable accuracy levels with minimal invasiveness, low complexity, and robustness to EM disturbances and harsh environmental conditions. The test case considered to illustrate the proposed methodology refers to a servomechanical application designed to monitor the health status in real-time of the flight control actuators using a model-based approach. Since the external aerodynamic loads acting on the system influence the operation of most of the actuators, their measurement would be helpful to accurately simulate the monitoring model's dynamic response. Therefore, the authors evaluate the proposed sensor fusion strategy effectiveness by using a distributed sensing of the airframe strain to infer the aerodynamic loads acting on the flight control actuator. Operationally speaking, a structural and an aerodynamic model are combined to generate a database used to train data-based surrogates correlating strain measurements to the corresponding actuator load.