Prognostics of onboard electrohydraulic servomechanisms: Proposal of a novel model-based fault detection neural technique

The paper aims to propose a new prognostic procedure centered on the characterization of the state of health of an electrohydraulic servomechanism typically used in aircraft primary flight controls. This approach is based on the innovative use of a model based fault detection and identification method (FDI) that applies artificial neural network in order to identify the actual state of wear of the actuator. The considered case study concerns an electrohydraulic actuator with a flapper-nozzle servo-valve whose progressive faults taken into consideration are: the clogging of the first stage of the flapper-nozzle valve, and the progressive gain loss of the torque motor. By means of the extraction of data from the system responses, under different extent of damage, multiple neural networks have been trained. Such networks have been integrated to create the prognostic method that has finally been tested under different conditions.